#!/usr/bin/env Rscript
# NOTE: The script now has 6 optional arguments:
#   1. Path to input easy results file
#   2. /output/ directory
#   3. Path to StudyInfo.csv
#   4. Path to sgd_gene_list
#   5. The experiment number (Exp# directory)
#   6. Standard deviation value

library("ggplot2")
library("plyr")
library("extrafont")
library("gridExtra")
library("gplots")
library("RColorBrewer")
library("stringr")
library("gdata")
library("plotly")
library("htmlwidgets")

# Parse arguments
args <- commandArgs(TRUE)
exp_number <- as.numeric(args[1])
delta_bg_factor <- as.numeric(args[2])
study_info_file <- file.path(args[3])
sgd_gene_list <- file.path(args[4])
input_file <- file.path(args[5])
out_dir <- file.path(args[6])

sprintf("The Standard Deviation value is: %f", delta_bg_factor)

out_dir_qc <- file.path(out_dir, "qc")

if (!dir.exists(out_dir)) {
  dir.create(out_dir)
}

if (!dir.exists(out_dir_qc)) {
  dir.create(out_dir_qc)
}

options(width = 1000)
ls.str()

# Write delBGFactor to the StudyInfo file
# TODO we probably shouldn't be doing this, need one source of truth
# TODO disabling this for now
# labels <- read.csv(file = study_info_file, stringsAsFactors = FALSE)  # sep = ","
# labels[exp_number, 3] <- delta_bg_factor
# write.csv(Labels, file = study_info_file, row.names = FALSE)

# Begin User Data Selection Section

# Read in the data
df <- read.delim(input_file, skip = 2, as.is = TRUE, row.names = 1, strip.white = TRUE)
df <- df[!(df[[1]] %in% c("", "Scan")), ]
# df <- df[!(df[[1]]%in%c(61:76)), ]  # remove dAmp plates which are Scans 61 thru 76
# df <- df[which(df$Specifics == "WT"), ]
# df_length <- length(df[1, ])
# df_end <- length(df[1, ]) - 2
# df <- df[, c(1:42, df_end:df_length)]

# print(names(df))

# Use numeric data to perform operations
df$col <- as.numeric(df$Col)
df$row <- as.numeric(df$Row)
df$l <- as.numeric(df$l)
df$k <- as.numeric(df$K)
df$r <- as.numeric(df$r)
df$scan <- as.numeric(df$Scan)
df$auc <- as.numeric(df$AUC)
df$last_bg <- as.numeric(df$LstBackgrd)
df$first_bg <- as.numeric(df$X1stBackgrd)

# print(df)

# Sometimes the non-varying drug is in the 'Drug' col vs the 'Modifier1' col
# as was the case in Gemcitabin and Cytarabin experiments.
# The following allows user to rename columns so as to get the appropriate
# data where it needs to be for the script to run properly.

# colnames(df)[7]  <- "Modifier1"
# colnames(df)[8]  <- "Conc1"
# colnames(df)[10]  <- "Drug"
# colnames(df)[11]  <- "Conc"

# Set the OrfRep to YDL227C for the ref data
df[df$ORF == "YDL227C", ]$OrfRep <- "YDL227C"

# Sean removes the Doxycyclin at 0.0ug.mL so that only the Oligomycin series with Doxycyclin of 0.12ug/mL are used.
# That is the first DM plates are removed from the data set with the following.
# This removes data with dox == 0 leaving gene expression on with four different concentrations of Gemcytabin
# df <- df[df$Conc1 != "0ug/ml", ]
df <- df[df$Drug != "BMH21", ] # this removes data concerning BMH21 for this experiment

# Mert placed the "bad_spot" text in the ORF col. for particular spots in the RF1 and RF2 plates.
# This code removes those spots from the data set used for the interaction analysis.
# Dr.Hartman feels that these donot effect Zscores significantly and so "non-curated" files were used.
# try(df <- df[df$ORF != "bad_spot", ])

# Get total number of drug concentrations
total_conc_nums <- length(unique(df$Conc))

# Function to ID numbers in string with characters+numbers (ie to get numeric drug conc)
numextract <- function(string) {
  str_extract(string, "\\-*\\d+\\.*\\d*")
}

# Generate a new column with the numeric drug concs
df$conc_num <- as.numeric(numextract(df$Conc))

# Generate new column with the numeric drug concs as factors starting at 0 for the graphing later
df$conc_num_factor <- as.numeric(as.factor(df$conc_num)) - 1

# Get the max factor for concentration
max_conc <- max(df$conc_num_factor)

# If treating numbers not as factors uncomment next line and comment out previous line
# max_conc <- max(df$conc_num)

# Remove wells with problems for making graphs and to not include in summary statistics
df <- df[df$Gene != "BLANK", ]
df <- df[df$Gene != "Blank", ]
df <- df[df$ORF != "Blank", ]
df <- df[df$Gene != "blank", ]
# df <- df[df$Gene != "HO", ]

# Use sgd_gene_list to update orfs and replace empty geneName cells with ORF name (adapted from Sean's Merge script).
# This is to 'fix' the naming for everything that follows (REMc, Heatmaps ... et.al) rather than do it piece meal later
# Sean's Match Script( which was adapted here) was fixed 2022_0608 so as not to overwrite the RF1&RF2 geneNames
# in the z_lm_l, k, r&auc output values. Values correlated well but were off by a multiplier factor.
genes <- data.frame(read.delim(
  file = sgd_gene_list, quote = "", header = FALSE, colClasses = c(rep("NULL", 3), rep("character", 2), rep("NULL", 11))))
for (i in 1:length(df[, 14])) {
  ii <- as.numeric(i)
  line_num <- match(df[ii, 14], genes[, 1], nomatch = 1)
  orf_rep_col_num <- as.numeric(match("OrfRep", names(df)))
  if (df[ii, orf_rep_col_num] != "YDL227C") {
    df[ii, 15] <- genes[line_num, 2]
  }
  if ((df[ii, 15] == "") || (df[ii, 15] == "OCT1")) {
    df[ii, 15] <- df[ii, orf_rep_col_num]
  }
}

# Remove dAmPs
# jlh confirmed to leave dAmps in so comment out this section
# DAmPs_list <- "../Code/22_0602_Remy_DAmPsList.txt"
# Damps <- read.delim(DAmPs_list, header = F)

# df <- df[!(df$ORF %in% Damps$V1), ]  # fix this to Damps[, 1]
# dfafterDampsRM = df  # backup for debugging

# Begin Graphics Boiler Plate Section
# theme elements for plots
theme_publication <- function(base_size = 14, base_family = "sans") {
  library(grid)
  library(ggthemes)
  (theme_foundation(base_size = base_size, base_family = base_family) +
    theme(plot.title = element_text(
      face = "bold",
      size = rel(1.2), hjust = 0.5),
      text = element_text(),
      panel.background = element_rect(colour = NA),
      plot.background = element_rect(colour = NA),
      panel.border = element_rect(colour = NA),
      axis.title = element_text(face = "bold", size = rel(1)),
      axis.title.y = element_text(angle = 90, vjust = 2),
      axis.title.x = element_text(vjust = -0.2),
      axis.text = element_text(),
      axis.line = element_line(colour = "black"),
      axis.ticks = element_line(),
      panel.grid.major = element_line(colour = "#f0f0f0"),
      panel.grid.minor = element_blank(),
      legend.key = element_rect(colour = NA),
      legend.position = "bottom",
      legend.direction = "horizontal",
      legend.key.size = unit(0.2, "cm"),
      legend.spacing = unit(0, "cm"),
      legend.title = element_text(face = "italic"),
      plot.margin = unit(c(10, 5, 5, 5), "mm"),
      strip.background = element_rect(colour = "#f0f0f0", fill = "#f0f0f0"),
      strip.text = element_text(face = "bold")
    )
  )
}

scale_fill_publication <- function(...) {
  library(scales)
  discrete_scale("fill", "Publication", manual_pal(
    values = c("#386cb0", "#fdb462", "#7fc97f", "#ef3b2c", "#662506", "#a6cee3", "#fb9a99", "#984ea3", "#ffff33")), ...)
}

scale_colour_publication <- function(...) {
  discrete_scale("colour", "Publication", manual_pal(
    values = c("#386cb0", "#fdb462", "#7fc97f", "#ef3b2c", "#662506", "#a6cee3", "#fb9a99", "#984ea3", "#ffff33")), ...)
}

theme_publication_legend_right <- function(base_size = 14, base_family = "sans") {
  (theme_foundation(base_size = base_size, base_family = base_family) +
    theme(plot.title = element_text(face = "bold", size = rel(1.2), hjust = 0.5),
      text = element_text(),
      panel.background = element_rect(colour = NA),
      plot.background = element_rect(colour = NA),
      panel.border = element_rect(colour = NA),
      axis.title = element_text(face = "bold", size = rel(1)),
      axis.title.y = element_text(angle = 90, vjust = 2),
      axis.title.x = element_text(vjust = -0.2),
      axis.text = element_text(),
      axis.line = element_line(colour = "black"),
      axis.ticks = element_line(),
      panel.grid.major = element_line(colour = "#f0f0f0"),
      panel.grid.minor = element_blank(),
      legend.key = element_rect(colour = NA),
      legend.position = "right",
      legend.direction = "vertical",
      legend.key.size = unit(0.5, "cm"),
      legend.spacing = unit(0, "cm"),
      legend.title = element_text(face = "italic"),
      plot.margin = unit(c(10, 5, 5, 5), "mm"),
      strip.background = element_rect(colour = "#f0f0f0", fill = "#f0f0f0"),
      strip.text = element_text(face = "bold")
    )
  )
}

scale_fill_publication <- function(...) {
  discrete_scale("fill", "Publication", manual_pal(
    values = c("#386cb0", "#fdb462", "#7fc97f", "#ef3b2c", "#662506", "#a6cee3", "#fb9a99", "#984ea3", "#ffff33")), ...)
}

scale_colour_publication <- function(...) {
  discrete_scale("colour", "Publication", manual_pal(
    values = c("#386cb0", "#fdb462", "#7fc97f", "#ef3b2c", "#662506", "#a6cee3", "#fb9a99", "#984ea3", "#ffff33")), ...)
}

# Print timestamp for initial time the code starts
timestamp()

# Begin QC Section
# Part 2 - Quality control
# Print quality control graphs for each dataset before removing contaminated data
# and before adjusting missing data to max theoretical values

# Plate analysis plot
# Plate analysis is a quality check to identify plate effects containing anomalies

plate_analysis_l <-
  ggplot(df, aes(Scan, l, color = as.factor(conc_num))) +
  geom_point(shape = 3, size = 0.2) +
  stat_summary(
    fun = mean,
    fun.min = function(x) mean(x) - sd(x),
    fun.max = function(x) mean(x) + sd(x),
    geom = "errorbar"
  ) +
  stat_summary(fun = mean, geom = "point", size = 0.6) +
  ggtitle("Plate analysis by Drug Conc for L before quality control") + theme_publication()

plate_analysis_k <-
  ggplot(df, aes(Scan, k, color = as.factor(conc_num))) +
  geom_point(shape = 3, size = 0.2) +
  stat_summary(
    fun = mean,
    fun.min = function(x) mean(x) - sd(x),
    fun.max = function(x) mean(x) + sd(x),
    geom = "errorbar"
  ) +
  stat_summary(fun = mean, geom = "point", size = 0.6) +
  ggtitle("Plate analysis by Drug Conc for K before quality control") + theme_publication()

plate_analysis_r <-
  ggplot(df, aes(Scan, r, color = as.factor(conc_num))) +
  geom_point(shape = 3, size = 0.2) +
  stat_summary(
    fun = mean,
    fun.min = function(x) mean(x) - sd(x),
    fun.max = function(x) mean(x) + sd(x),
    geom = "errorbar"
  ) +
  stat_summary(fun = mean, geom = "point", size = 0.6) +
  ggtitle("Plate analysis by Drug Conc for r before quality control") + theme_publication()

plate_analysis_auc <-
  ggplot(df, aes(Scan, auc, color = as.factor(conc_num))) +
  geom_point(shape = 3, size = 0.2) +
  stat_summary(
    fun = mean,
    fun.min = function(x) mean(x) - sd(x),
    fun.max = function(x) mean(x) + sd(x),
    geom = "errorbar"
  ) +
  stat_summary(fun = mean, geom = "point", size = 0.6) +
  ggtitle("Plate analysis by Drug Conc for auc before quality control") + theme_publication()

plate_analysis_l_box <-
  ggplot(df, aes(as.factor(Scan), l, color = as.factor(conc_num))) +
  geom_boxplot() +
  ggtitle("Plate analysis by Drug Conc for L before quality control") +
  theme_publication()

plate_analysis_k_box <-
  ggplot(df, aes(as.factor(Scan), k, color = as.factor(conc_num))) +
  geom_boxplot() +
  ggtitle("Plate analysis by Drug Conc for K before quality control") +
  theme_publication()

plate_analysis_r_box <-
  ggplot(df, aes(as.factor(Scan), r, color = as.factor(conc_num))) +
  geom_boxplot() +
  ggtitle("Plate analysis by Drug Conc for r before quality control") +
  theme_publication()

plate_analysis_auc_box <-
  ggplot(df, aes(as.factor(Scan), auc, color = as.factor(conc_num))) +
  geom_boxplot() +
  ggtitle("Plate analysis by Drug Conc for auc before quality control") +
  theme_publication()

# Quality control - values with a high delta background likely have heavy contamination
# Check the frequency of these values
# Report the L and k values of these spots
# Report the number to be removed based on the delta_background_tolerance
df$delta_bg <- df$last_bg - df$first_bg

# Raw l vs k before QC
raw_l_vs_k_before_qc <-
  ggplot(df, aes(l, k, color = as.factor(conc_num))) +
  geom_point(aes(ORF = ORF, Gene = Gene, delta_bg = delta_bg), shape = 3) +
  ggtitle("Raw L vs K before QC") +
  theme_publication_legend_right()

pdf(file.path(out_dir_qc, "raw_l_vs_k_before_qc.pdf"), width = 12, height = 8)
raw_l_vs_k_before_qc

dev.off()

pgg <- ggplotly(raw_l_vs_k_before_qc)
plotly_path <- file.path(out_dir_qc, "raw_l_vs_k_before_qc.html")
saveWidget(pgg, file = plotly_path, selfcontained = TRUE)

# Set delta background tolerance based on 3 sds from the mean delta background
delta_background_tolerance <- mean(df$delta_bg) + (delta_bg_factor * sd(df$delta_bg))
# delta_background_tolerance <- mean(df$delta_bg)+(3*sd(df$delta_bg))
sprintf("delta_background_tolerance is %f", delta_background_tolerance)

plate_analysis_delta_bg <-
  ggplot(df, aes(Scan, delta_bg, color = as.factor(conc_num))) +
  geom_point(shape = 3, size = 0.2, position = "jitter") +
  stat_summary(
    fun = mean,
    fun.min = function(x) mean(x) - sd(x),
    fun.max = function(x) mean(x) + sd(x),
    geom = "errorbar"
  ) +
  stat_summary(fun = mean, geom = "point", size = 0.6) +
  ggtitle("Plate analysis by Drug Conc for delta_bg before quality control") +
  theme_publication()

plate_analysis_delta_bg_box <-
  ggplot(df, aes(as.factor(Scan), delta_bg, color = as.factor(conc_num))) +
  geom_boxplot() +
  ggtitle("Plate analysis by Drug Conc for delta_bg before quality control") +
  theme_publication()

x_delta_bg_above_tolerance <-
  df[df$delta_bg >= delta_background_tolerance, ]
x_delta_bg_above_tolerance_k_halfmedian <-
  (median(x_delta_bg_above_tolerance$k, na.rm = TRUE)) / 2
x_delta_bg_above_tolerance_l_halfmedian <-
  (median(x_delta_bg_above_tolerance$l, na.rm = TRUE)) / 2
x_delta_bg_above_tolerance_to_remove <-
  dim(x_delta_bg_above_tolerance)[1]

x_delta_bg_above_tolerance_l_vs_k <-
  ggplot(x_delta_bg_above_tolerance, aes(l, k, color = as.factor(conc_num))) +
  geom_point(aes(ORF = ORF, Gene = Gene, delta_bg = delta_bg), shape = 3) +
  ggtitle(paste("Raw L vs K for strains above delta background threshold of", delta_background_tolerance, "or above")) +
  annotate("text", x = x_delta_bg_above_tolerance_l_halfmedian, y = x_delta_bg_above_tolerance_k_halfmedian,
    label = paste("Strains above delta background tolerance = ", x_delta_bg_above_tolerance_to_remove)
  ) +
  theme_publication_legend_right()

pdf(file.path(out_dir_qc, "raw_l_vs_k_for_strains_above_delta_background_threshold.pdf"), width = 12, height = 8)
x_delta_bg_above_tolerance_l_vs_k

dev.off()

pgg <- ggplotly(x_delta_bg_above_tolerance_l_vs_k)
plotly_path <- file.path(out_dir_qc, "raw_l_vs_k_for_strains_above_delta_background_threshold.html")
saveWidget(pgg, file = plotly_path, selfcontained = TRUE)

# Frequency plot for all data vs. the delta_background
delta_bg_frequency_plot <- ggplot(df, aes(delta_bg, color = as.factor(conc_num))) + geom_density() +
  ggtitle("Density plot for Delta Background by Conc All Data") + theme_publication_legend_right()

# Bar plot for all data vs. the delta_background
delta_bg_bar_plot <- ggplot(df, aes(delta_bg, color = as.factor(conc_num))) + geom_bar() +
  ggtitle("Bar plot for Delta Background by Conc All Data") + theme_publication_legend_right()

pdf(file.path(out_dir_qc, "frequency_delta_background.pdf"), width = 12, height = 8)
print(delta_bg_frequency_plot)
print(delta_bg_bar_plot)
dev.off()

# Need to identify missing data, and differentiate between this data and removed data
# so the removed data can get set to NA and the missing data can get set to max theoretical values
# 1 for missing data, 0 for non missing data
# Use "NG" for NoGrowth rather than "missing"
df$NG <- 0
try(df[df$l == 0 & !is.na(df$l), ]$NG <- 1)

# 1 for removed data, 0 non removed data
# Use DB to identify number of genes removed due to the DeltaBackground Threshold rather than "Removed"
df$DB <- 0
try(df[df$delta_bg >= delta_background_tolerance, ]$DB <- 1)

# Replace the CPPs for l, r, auc and k (must be last!) for removed data
try(df[df$delta_bg >= delta_background_tolerance, ]$l <- NA)
try(df[df$delta_bg >= delta_background_tolerance, ]$r <- NA)
try(df[df$delta_bg >= delta_background_tolerance, ]$auc <- NA)
try(df[df$delta_bg >= delta_background_tolerance, ]$k <- NA)

# QC Plots
plate_analysis_l_after_qc <- ggplot(df, aes(Scan, l, color = as.factor(conc_num))) + geom_point(shape = 3, size = 0.2) +
  stat_summary(
    fun = mean,
    fun.min = function(x) mean(x) - sd(x),
    fun.max = function(x) mean(x) + sd(x),
    geom = "errorbar"
  ) +
  stat_summary(
    fun = mean,
    geom = "point",
    size = 0.6
  ) +
  ggtitle("Plate analysis by Drug Conc for L after quality control") + theme_publication()

plate_analysis_k_after_qc <-
  ggplot(df, aes(Scan, k, color = as.factor(conc_num))) +
  geom_point(shape = 3, size = 0.2) +
  stat_summary(
    fun = mean,
    fun.min = function(x) mean(x) - sd(x),
    fun.max = function(x) mean(x) + sd(x),
    geom = "errorbar"
  ) +
  stat_summary(
    fun = mean,
    geom = "point",
    size = 0.6
  ) +
  ggtitle("Plate analysis by Drug Conc for k after quality control") + theme_publication()

plate_analysis_r_after_qc <-
  ggplot(df, aes(Scan, r, color = as.factor(conc_num))) +
  geom_point(shape = 3, size = 0.2) +
  stat_summary(
    fun = mean,
    fun.min = function(x) mean(x) - sd(x),
    fun.max = function(x) mean(x) + sd(x),
    geom = "errorbar"
  ) +
  stat_summary(
    fun = mean,
    geom = "point",
    size = 0.6
  ) +
  ggtitle("Plate analysis by Drug Conc for r after quality control") + theme_publication()

plate_analysis_auc_after_qc <-
  ggplot(df, aes(Scan, auc, color = as.factor(conc_num))) +
  geom_point(shape = 3, size = 0.2) +
  stat_summary(
    fun = mean,
    fun.min = function(x) mean(x) - sd(x),
    fun.max = function(x) mean(x) + sd(x),
    geom = "errorbar"
  ) +
  stat_summary(
    fun = mean,
    geom = "point",
    size = 0.6
  ) +
  ggtitle("Plate analysis by Drug Conc for auc after quality control") + theme_publication()

plate_analysis_delta_bg_after_qc <-
  ggplot(df, aes(Scan, delta_bg, color = as.factor(conc_num))) +
  geom_point(shape = 3, size = 0.2) +
  stat_summary(
    fun = mean,
    fun.min = function(x) mean(x) - sd(x),
    fun.max = function(x) mean(x) + sd(x),
    geom = "errorbar"
  ) +
  stat_summary(
    fun = mean,
    geom = "point",
    size = 0.6
  ) +
  ggtitle("Plate analysis by Drug Conc for delta_bg after quality control") +
  theme_publication()

plate_analysis_l_box_after_qc <-
  ggplot(df, aes(as.factor(Scan), l, color = as.factor(conc_num))) +
  geom_boxplot() +
  ggtitle("Plate analysis by Drug Conc for L after quality control") +
  theme_publication()

plate_analysis_k_box_after_qc <-
  ggplot(df, aes(as.factor(Scan), k, color = as.factor(conc_num))) +
  geom_boxplot() +
  ggtitle("Plate analysis by Drug Conc for K after quality control") +
  theme_publication()

plate_analysis_r_box_after_qc <-
  ggplot(df, aes(as.factor(Scan), r, color = as.factor(conc_num))) +
  geom_boxplot() +
  ggtitle("Plate analysis by Drug Conc for r after quality control") +
  theme_publication()

plate_analysis_auc_box_after_qc <-
  ggplot(df, aes(as.factor(Scan), auc, color = as.factor(conc_num))) +
  geom_boxplot() +
  ggtitle("Plate analysis by Drug Conc for auc after quality control") +
  theme_publication()

plate_analysis_delta_bg_box_after_qc <-
  ggplot(df, aes(as.factor(Scan), delta_bg, color = as.factor(conc_num))) +
  geom_boxplot() +
  ggtitle("Plate analysis by Drug Conc for delta_bg after quality control") +
  theme_publication()

# Print the plate analysis data before and after QC
pdf(file.path(out_dir_qc, "plate_analysis.pdf"), width = 14, height = 9)
plate_analysis_l
plate_analysis_l_after_qc
plate_analysis_k
plate_analysis_k_after_qc
plate_analysis_r
plate_analysis_r_after_qc
plate_analysis_auc
plate_analysis_auc_after_qc
plate_analysis_delta_bg
plate_analysis_delta_bg_after_qc
dev.off()

# Print the plate analysis data before and after QC
pdf(file.path(out_dir_qc, "plate_analysis_boxplots.pdf"), width = 18, height = 9)
plate_analysis_l_box
plate_analysis_l_box_after_qc
plate_analysis_k_box
plate_analysis_k_box_after_qc
plate_analysis_r_box
plate_analysis_r_box_after_qc
plate_analysis_auc_box
plate_analysis_auc_box_after_qc
plate_analysis_delta_bg_box
plate_analysis_delta_bg_box_after_qc
dev.off()

# Remove the zero values and print plate analysis
x_no_zero <- df[which(df$l > 0), ]
plate_analysis_l_after_qc_z <-
  ggplot(x_no_zero, aes(Scan, l, color = as.factor(conc_num))) +
  geom_point(shape = 3, size = 0.2) +
  stat_summary(
    fun = mean,
    fun.min = function(x) mean(x) - sd(x),
    fun.max = function(x) mean(x) + sd(x),
    geom = "errorbar"
  ) +
  stat_summary(
    fun = mean,
    geom = "point",
    size = 0.6
  ) +
  ggtitle("Plate analysis by Drug Conc for L after quality control") +
  theme_publication()

plate_analysis_k_after_qc_z <-
  ggplot(x_no_zero, aes(Scan, k, color = as.factor(conc_num))) +
  geom_point(shape = 3, size = 0.2) +
  stat_summary(
    fun = mean,
    fun.min = function(x) mean(x) - sd(x),
    fun.max = function(x) mean(x) + sd(x),
    geom = "errorbar"
  ) +
  stat_summary(
    fun = mean,
    geom = "point",
    size = 0.6
  ) +
  ggtitle("Plate analysis by Drug Conc for K after quality control") +
  theme_publication()

plate_analysis_r_after_qc_z <-
  ggplot(x_no_zero, aes(Scan, r, color = as.factor(conc_num))) +
  geom_point(shape = 3, size = 0.2) +
  stat_summary(
    fun = mean,
    fun.min = function(x) mean(x) - sd(x),
    fun.max = function(x) mean(x) + sd(x),
    geom = "errorbar"
  ) +
  stat_summary(
    fun = mean,
    geom = "point",
    size = 0.6
  ) +
  ggtitle("Plate analysis by Drug Conc for r after quality control") +
  theme_publication()

plate_analysis_auc_after_qc_z <-
  ggplot(x_no_zero, aes(Scan, auc, color = as.factor(conc_num))) +
  geom_point(shape = 3, size = 0.2) +
  stat_summary(
    fun = mean,
    fun.min = function(x) mean(x) - sd(x),
    fun.max = function(x) mean(x) + sd(x),
    geom = "errorbar"
  ) +
  stat_summary(
    fun = mean,
    geom = "point",
    size = 0.6
  ) +
  ggtitle("Plate analysis by Drug Conc for auc after quality control") +
  theme_publication()

plate_analysis_delta_bg_after_qc_z <-
  ggplot(x_no_zero, aes(Scan, delta_bg, color = as.factor(conc_num))) +
  geom_point(shape = 3, size = 0.2) +
  stat_summary(
    fun = mean,
    fun.min = function(x) mean(x) - sd(x),
    fun.max = function(x) mean(x) + sd(x),
    geom = "errorbar"
  ) +
  stat_summary(
    fun = mean,
    geom = "point",
    size = 0.6
  ) +
  ggtitle("Plate analysis by Drug Conc for delta_bg after quality control") +
  theme_publication()

plate_analysis_l_box_after_qc_z <-
  ggplot(x_no_zero, aes(as.factor(Scan), l, color = as.factor(conc_num))) +
  geom_boxplot() +
  ggtitle("Plate analysis by Drug Conc for L after quality control") +
  theme_publication()

plate_analysis_k_box_after_qc_z <-
  ggplot(x_no_zero, aes(as.factor(Scan), k, color = as.factor(conc_num))) +
  geom_boxplot() +
  ggtitle("Plate analysis by Drug Conc for K after quality control") +
  theme_publication()

plate_analysis_r_box_after_qc_z <-
  ggplot(x_no_zero, aes(as.factor(Scan), r, color = as.factor(conc_num))) +
  geom_boxplot() +
  ggtitle("Plate analysis by Drug Conc for r after quality control") +
  theme_publication()

plate_analysis_auc_box_after_qc_z <-
  ggplot(x_no_zero, aes(as.factor(Scan), auc, color = as.factor(conc_num))) +
  geom_boxplot() +
  ggtitle("Plate analysis by Drug Conc for auc after quality control") +
  theme_publication()

plate_analysis_delta_bg_box_after_qc_z <-
  ggplot(x_no_zero, aes(as.factor(Scan), delta_bg, color = as.factor(conc_num))) +
  geom_boxplot() +
  ggtitle("Plate analysis by Drug Conc for delta_bg after quality control") +
  theme_publication()

# Print the plate analysis data before and after QC
pdf(file.path(out_dir_qc, "plate_analysis_no_zeros.pdf"), width = 14, height = 9)
plate_analysis_l_after_qc_z
plate_analysis_k_after_qc_z
plate_analysis_r_after_qc_z
plate_analysis_auc_after_qc_z
plate_analysis_delta_bg_after_qc_z
dev.off()

# Print the plate analysis data before and after QC
pdf(file.path(out_dir_qc, "plate_analysis_no_zeros_boxplots.pdf"), width = 18, height = 9)
plate_analysis_l_box_after_qc_z
plate_analysis_k_box_after_qc_z
plate_analysis_r_box_after_qc_z
plate_analysis_auc_box_after_qc_z
plate_analysis_delta_bg_box_after_qc_z
dev.off()

# Remove dataset with zeros removed
rm(x_no_zero)

# df_test_missing_and_removed <- df[df$Removed == 1, ]

# Calculate summary statistics for all strains, including both background and the deletions
x_stats_all <- ddply(
  df,
  c("conc_num", "conc_num_factor"),
  summarise,
  N = (length(l)),
  mean_l = mean(l, na.rm = TRUE),
  median_l = median(l, na.rm = TRUE),
  max_l = max(l, na.rm = TRUE),
  min_l = min(l, na.rm = TRUE),
  sd_l = sd(l, na.rm = TRUE),
  se_l = sd_l / sqrt(N - 1),
  mean_k = mean(k, na.rm = TRUE),
  median_k = median(k, na.rm = TRUE),
  max_k = max(k, na.rm = TRUE),
  min_k = min(k, na.rm = TRUE),
  sd_k = sd(k, na.rm = TRUE),
  se_k = sd_k / sqrt(N - 1),
  mean_r = mean(r, na.rm = TRUE),
  median_r = median(r, na.rm = TRUE),
  max_r = max(r, na.rm = TRUE),
  min_r = min(r, na.rm = TRUE),
  sd_r = sd(r, na.rm = TRUE),
  se_r = sd_r / sqrt(N - 1),
  mean_auc = mean(auc, na.rm = TRUE),
  median_auc = median(auc, na.rm = TRUE),
  max_auc = max(auc, na.rm = TRUE),
  min_auc = min(auc, na.rm = TRUE),
  sd_auc = sd(auc, na.rm = TRUE),
  se_auc = sd_auc / sqrt(N - 1)
)

# print(x_stats_all_l)
write.csv(x_stats_all, file.path(out_dir, "summary_stats_all_strains.csv"), row.names = FALSE)

# Part 3 - Generate summary statistics and calculate the max theoretical L value
# Calculate the Z score at each drug conc for each deletion strain

# Get the background strains - can be modified to take another argument but for most screens will just be YDL227C
background_strains <- c("YDL227C")

# First part of loop will go through for each background strain
# In most cases there will only be one YDL227C
for (s in background_strains) {
  x_background <- df[df$OrfRep == s, ]

  # If there's missing data for the background strains set these values to NA so the 0 values aren't included in summary statistics
  # we may want to consider in some cases giving the max high value to L depending on the data type
  if (table(x_background$l)[1] == 0) {
    x_background[x_background$l == 0, ]$l <- NA
    x_background[x_background$k == 0, ]$k <- NA
    x_background[x_background$r == 0, ]$r <- NA
    x_background[x_background$auc == 0, ]$auc <- NA
  }

  x_background <- x_background[!is.na(x_background$l), ]

  # Get summary stats for L, k, R, auc
  x_stats_by_l <- ddply(
    x_background,
    c("OrfRep", "conc_num", "conc_num_factor"),
    summarise,
    N = (length(l)),
    mean = mean(l, na.rm = TRUE),
    median = median(l, na.rm = TRUE),
    max = max(l, na.rm = TRUE),
    min = min(l, na.rm = TRUE),
    sd = sd(l, na.rm = TRUE),
    se = sd / sqrt(N - 1)
  )

  print(x_stats_by_l)
  x1_sd <- max(x_stats_by_l$sd)

  x_stats_by_k <- ddply(
    x_background,
    c("OrfRep", "conc_num", "conc_num_factor"),
    summarise,
    N = (length(k)),
    mean = mean(k, na.rm = TRUE),
    median = median(k, na.rm = TRUE),
    max = max(k, na.rm = TRUE),
    min = min(k, na.rm = TRUE),
    sd = sd(k, na.rm = TRUE),
    se = sd / sqrt(N - 1)
  )
  
  x1_sd_k <- max(x_stats_by_k$sd)
  
  x_stats_by_r <- ddply(
    x_background,
    c("OrfRep", "conc_num", "conc_num_factor"),
    summarise,
    N = length(r),
    mean = mean(r, na.rm = TRUE),
    median = median(r, na.rm = TRUE),
    max = max(r, na.rm = TRUE),
    min = min(r, na.rm = TRUE),
    sd = sd(r, na.rm = TRUE),
    se = sd / sqrt(N - 1)
  )
  
  x1_sd_r <- max(x_stats_by_r$sd)
  
  x_stats_by_auc <- ddply(
    x_background,
    c("OrfRep", "conc_num", "conc_num_factor"),
    summarise,
    N = length(auc),
    mean = mean(auc, na.rm = TRUE),
    median = median(auc, na.rm = TRUE),
    max = max(auc, na.rm = TRUE),
    min = min(auc, na.rm = TRUE),
    sd = sd(auc, na.rm = TRUE),
    se = sd / sqrt(N - 1)
  )
  
  x1_sd_auc <- max(x_stats_by_auc$sd)
  
  x_stats_by <- ddply(
    x_background,
    c("OrfRep", "conc_num", "conc_num_factor"),
    summarise,
    N = (length(l)),
    mean_l = mean(l, na.rm = TRUE),
    median_l = median(l, na.rm = TRUE),
    max_l = max(l, na.rm = TRUE),
    min_l = min(l, na.rm = TRUE),
    sd_l = sd(l, na.rm = TRUE),
    se_l = sd_l / sqrt(N - 1),
    mean_k = mean(k, na.rm = TRUE),
    median_k = median(k, na.rm = TRUE),
    max_k = max(k, na.rm = TRUE),
    min_k = min(k, na.rm = TRUE),
    sd_k = sd(k, na.rm = TRUE),
    se_k = sd_k / sqrt(N - 1),
    mean_r = mean(r, na.rm = TRUE),
    median_r = median(r, na.rm = TRUE),
    max_r = max(r, na.rm = TRUE),
    min_r = min(r, na.rm = TRUE),
    sd_r = sd(r, na.rm = TRUE),
    se_r = sd_r / sqrt(N - 1),
    mean_auc = mean(auc, na.rm = TRUE),
    median_auc = median(auc, na.rm = TRUE),
    max_auc = max(auc, na.rm = TRUE),
    min_l = min(auc, na.rm = TRUE),
    sd_auc = sd(auc, na.rm = TRUE),
    se_auc = sd_auc / sqrt(N - 1)
  )
  
  write.csv(x_stats_by, file.path(out_dir, "summary_stats_background_strains.csv"), row.names = FALSE)
  
  # Calculate the max theoretical L values
  # Only look for max values when k is within 2sd of the ref strain
  for (q in unique(df$conc_num_factor)) {
    if (q == 0) {
      x_within_2sd_k <-
        df[df$conc_num_factor == q, ]
      x_within_2sd_k <-
        x_within_2sd_k[!is.na(x_within_2sd_k$l), ]
      x_stats_temp_k <-
        x_stats_by_k[x_stats_by_k$conc_num_factor == q, ]
      x_within_2sd_k <-
        x_within_2sd_k[x_within_2sd_k$k >= (x_stats_temp_k$mean[1] - (2 * x_stats_temp_k$sd[1])), ]
      x_within_2sd_k <-
        x_within_2sd_k[x_within_2sd_k$k <= (x_stats_temp_k$mean[1] + (2 * x_stats_temp_k$sd[1])), ]
      x_outside_2sd_k <-
        df[df$conc_num_factor == q, ]
      x_outside_2sd_k <-
        x_outside_2sd_k[!is.na(x_outside_2sd_k$l), ]
      # x_outside_2sd_k_temp <-
      #   x_stats_by_k[x_stats_by_k$conc_num_factor == q, ]
      x_outside_2sd_k <-
        x_outside_2sd_k[
          x_outside_2sd_k$k <= (x_stats_temp_k$mean[1] - (2 * x_stats_temp_k$sd[1])) |
          x_outside_2sd_k$k >= (x_stats_temp_k$mean[1] + (2 * x_stats_temp_k$sd[1])), ]
      # x_outside_2sd_k <-
      #   x_outside_2sd_k[x_outside_2sd_k$k >= (x_stats_temp_k$mean[1] + (2*x_stats_temp_k$sd[1])), ]
    }
    if (q > 0) {
      x_within_2sd_k_temp <-
        df[df$conc_num_factor == q, ]
      x_within_2sd_k_temp <-
        x_within_2sd_k_temp[!is.na(x_within_2sd_k_temp$l), ]
      x_stats_temp_k <-
        x_stats_by_k[x_stats_by_k$conc_num_factor == q, ]
      x_within_2sd_k_temp <-
        x_within_2sd_k_temp[x_within_2sd_k_temp$k >= (x_stats_temp_k$mean[1] - (2 * x_stats_temp_k$sd[1])), ]
      x_within_2sd_k_temp <-
        x_within_2sd_k_temp[x_within_2sd_k_temp$k <= (x_stats_temp_k$mean[1] + (2 * x_stats_temp_k$sd[1])), ]
      x_within_2sd_k <-
        rbind(x_within_2sd_k, x_within_2sd_k_temp)
      x_outside_2sd_k_temp <-
        df[df$conc_num_factor == q, ]
      x_outside_2sd_k_temp <-
        x_outside_2sd_k_temp[!is.na(x_outside_2sd_k_temp$l), ]
      # x_outside_2sd_k_temp <-
      #   x_stats_by_k[x_stats_by_k$conc_num_factor == q, ]
      x_outside_2sd_k_temp <-
        x_outside_2sd_k_temp[
          x_outside_2sd_k_temp$k <= (x_stats_temp_k$mean[1] - (2 * x_stats_temp_k$sd[1])) |
          x_outside_2sd_k_temp$k >= (x_stats_temp_k$mean[1] + (2 * x_stats_temp_k$sd[1])), ]
      # x_outside_2sd_k_temp <-
      #   x_outside_2sd_k_temp[x_outside_2sd_k_temp$k >= (x_stats_temp_k$mean[1] + (2*x_stats_temp_k$sd[1])) , ]
      x_outside_2sd_k <-
        rbind(x_outside_2sd_k, x_outside_2sd_k_temp)
    }
  }

  x_stats_by_l_within_2sd_k <- ddply(
    x_within_2sd_k,
    c("conc_num", "conc_num_factor"),
    summarise,
    N = (length(l)),
    mean = mean(l),
    median = median(l),
    max = max(l, na.rm = TRUE),
    min = min(l, na.rm = TRUE),
    sd = sd(l),
    se = sd / sqrt(N - 1),
    z_max = (max - mean) / sd
  )

  print(x_stats_by_l_within_2sd_k)

  x1_sd_within_2sd_k <- max(x_stats_by_l_within_2sd_k$sd)

  write.csv(
    x_stats_by_l_within_2sd_k,
    file.path(out_dir_qc, "max_observed_l_vals_for_spots_within_2sd_k.csv"),
    row.names = FALSE
  )
  
  x_stats_by_l_outside_2sd_k <- ddply(
    x_outside_2sd_k,
    c("conc_num", "conc_num_factor"),
    summarise,
    N = (length(l)),
    mean = mean(l),
    median = median(l),
    max = max(l, na.rm = TRUE),
    min = min(l, na.rm = TRUE),
    sd = sd(l),
    se = sd / sqrt(N - 1)
  )

  print(x_stats_by_l_outside_2sd_k)
  x1_sd_outside_2sd_k <- max(x_stats_by_l_outside_2sd_k$sd)

  # x1_sd_outside_2sd_k <- df[df$l %in% x1_sd_within_2sd_k$l, ]
  outside_2sd_k_l_vs_k <-
    ggplot(x_outside_2sd_k, aes(l, k, color = as.factor(conc_num))) +
    geom_point(aes(ORF = ORF, Gene = Gene, delta_bg = delta_bg), shape = 3) +
    ggtitle("Raw L vs K for strains falling outside 2sd of the K mean at each conc") +
    theme_publication_legend_right()
  
  pdf(file.path(out_dir_qc, "raw_l_vs_k_for_strains_2sd_outside_mean_k.pdf"), width = 10, height = 8)
  print(outside_2sd_k_l_vs_k)

  dev.off()

  pgg <- ggplotly(outside_2sd_k_l_vs_k)
  plotly_path <- file.path(out_dir_qc, "raw_l_vs_k_for_strains_outside_2sd_k.html")
  saveWidget(pgg, file = plotly_path, selfcontained = TRUE)

  outside_2sd_k_delta_background_vs_k <-
    ggplot(x_outside_2sd_k, aes(delta_bg, k, color = as.factor(conc_num))) +
    geom_point(aes(l = l, ORF = ORF, Gene = Gene), shape = 3, position = "jitter") +
    ggtitle("DeltaBackground vs K for strains falling outside 2sd of the K mean at each conc") +
    theme_publication_legend_right()

  pdf(file.path(out_dir_qc, "delta_background_vs_k_for_strains_2sd_outside_mean_k.pdf"), width = 10, height = 8)
  outside_2sd_k_delta_background_vs_k
  dev.off()
  pgg <- ggplotly(outside_2sd_k_delta_background_vs_k)

  # pgg
  plotly_path <- file.path(out_dir_qc, "delta_background_vs_k_for_strains_outside_2sd_k.html")
  saveWidget(pgg, file = plotly_path, selfcontained = TRUE)

  # Get the background strain mean values at the no drug conc to calculate shift
  background_l <- x_stats_by_l$mean[1]
  background_k <- x_stats_by_k$mean[1]
  background_r <- x_stats_by_r$mean[1]
  background_auc <- x_stats_by_auc$mean[1]

  # Create empty plots for plotting element
  p_l <- ggplot()
  p_k <- ggplot()
  p_r <- ggplot()
  p_auc <- ggplot()

  p_rf_l <- ggplot()
  p_rf_k <- ggplot()
  p_rf_r <- ggplot()
  p_rf_auc <- ggplot()

  # Get only the deletion strains
  df2 <- df
  df2 <- df2[df2$OrfRep != "YDL227C", ]

  # If set to max theoretical value, add a 1 to SM, if not, leave as 0
  # SM = Set to Max
  df2$SM <- 0
  # Set the missing values to the highest theoretical value at each drug conc for L
  # Leave other values as 0 for the max/min
  for (i in 1:length(unique(df2$conc_num))) {
    concentration <- unique(df2$conc_num)[i]
    df2_temp <- df2[df2$conc_num == concentration, ]
    if (concentration == 0) {
      df2_new <- df2_temp
      sprintf("Check loop order, conc = %f", concentration)
    }
    if (concentration > 0) {
      try(df2_temp[df2_temp$l == 0 & !is.na(df2_temp$l), ]$l <- x_stats_by_l_within_2sd_k$max[i])
      try(df2_temp[df2_temp$l >= x_stats_by_l_within_2sd_k$max[i] & !is.na(df2_temp$l), ]$SM <- 1)
      try(df2_temp[df2_temp$l >= x_stats_by_l_within_2sd_k$max[i] & !is.na(df2_temp$l), ]$l <- x_stats_by_l_within_2sd_k$max[i])
      # df2_temp[df2_temp$k == 0, ]$k <- x_stats_all_k$max[i]
      # df2_temp[df2_temp$r == 0, ]$r <- x_stats_all_r$max[i]
      # df2_temp[df2_temp$auc == 0, ]$auc <- x_stats_all_auc$max[i]
      sprintf("Check loop order, conc = %f", concentration)
      df2_new <- rbind(df2_new, df2_temp)
    }
  }

  df2 <- df2_new

  # Get only the RF strains
  df2_rf <- df
  df2_rf <- df2_rf[df2_rf$OrfRep == "YDL227C", ]
  # If set to max theoretical value, add a 1 to SM, if not, leave as 0
  # SM = Set to Max
  df2_rf$SM <- 0
  # Set the missing values to the highest theoretical value at each drug conc for L
  # Leave other values as 0 for the max/min
  for (i in 1:length(unique(df2_rf$conc_num))) {
    concentration <- unique(df2_rf$conc_num)[i]
    df2_rf_temp <- df2_rf[df2_rf$conc_num == concentration, ]
    if (concentration == 0) {
      df2_rf_new <- df2_rf_temp
      sprintf("Check loop order, conc = %f", concentration)
    }
    if (concentration > 0) {
      try(df2_rf_temp[df2_rf_temp$l == 0 & !is.na(df2_rf_temp$l), ]$l <- x_stats_by_l_within_2sd_k$max[i])
      try(df2_temp[df2_temp$l >= x_stats_by_l_within_2sd_k$max[i] & !is.na(df2_temp$l), ]$SM <- 1)
      try(df2_rf_temp[df2_rf_temp$l >= x_stats_by_l_within_2sd_k$max[i] & !is.na(df2_rf_temp$l), ]$l <- x_stats_by_l_within_2sd_k$max[i])
      sprintf("If error, refs have no L values outside theoretical max L, for REFs, conc = %f", concentration)
      df2_rf_new <- rbind(df2_rf_new, df2_rf_temp)
    }
  }
  df2_rf <- df2_rf_new

  # Part 4 Get the RF Z score values
  # Change the OrfRep Column to include the RF strain, the Gene name and the Num. so each RF gets its own score
  df2_rf$OrfRep <- paste(df2_rf$OrfRep, df2_rf$Gene, df2_rf$Num., sep = "_")

  num_genes_rf <- length(unique(df2_rf$OrfRep))
  # print(num_genes_rf)

  # Create the output data.frame containing columns for each RF strain
  interaction_scores_rf <- unique(df2_rf["OrfRep"])
  # interaction_scores_rf$Gene <- unique(df2$Gene)
  interaction_scores_rf$Gene <- NA
  interaction_scores_rf$Raw_Shift_l <- NA
  interaction_scores_rf$z_shift_l <- NA
  interaction_scores_rf$lm_Score_l <- NA
  interaction_scores_rf$z_lm_l <- NA
  interaction_scores_rf$R_Squared_l <- NA
  interaction_scores_rf$Sum_z_Score_l <- NA
  interaction_scores_rf$avg_zscore_l <- NA
  interaction_scores_rf$Raw_Shift_k <- NA
  interaction_scores_rf$z_shift_k <- NA
  interaction_scores_rf$lm_Score_k <- NA
  interaction_scores_rf$z_lm_k <- NA
  interaction_scores_rf$R_Squared_k <- NA
  interaction_scores_rf$Sum_z_Score_k <- NA
  interaction_scores_rf$avg_zscore_k <- NA
  interaction_scores_rf$Raw_Shift_r <- NA
  interaction_scores_rf$z_shift_r <- NA
  interaction_scores_rf$lm_Score_r <- NA
  interaction_scores_rf$z_lm_r <- NA
  interaction_scores_rf$R_Squared_r <- NA
  interaction_scores_rf$Sum_z_Score_r <- NA
  interaction_scores_rf$avg_zscore_r <- NA
  interaction_scores_rf$Raw_Shift_auc <- NA
  interaction_scores_rf$z_shift_auc <- NA
  interaction_scores_rf$lm_Score_auc <- NA
  interaction_scores_rf$z_lm_auc <- NA
  interaction_scores_rf$R_Squared_auc <- NA
  interaction_scores_rf$Sum_z_Score_auc <- NA
  interaction_scores_rf$avg_zscore_auc <- NA
  interaction_scores_rf$NG <- NA
  interaction_scores_rf$SM <- NA

  for (i in 1:num_genes_rf) {
    # Get each deletion strain ORF
    gene_sel <- unique(df2_rf$OrfRep)[i]
    # Extract only the current deletion strain and its data
    x_gene_sel <- df2_rf[df2_rf$OrfRep == gene_sel, ]

    x_stats_interaction <- ddply(
      x_gene_sel,
      c("OrfRep", "Gene", "conc_num", "conc_num_factor"),
      summarise,
      N = (length(l)),
      mean_l = mean(l, na.rm = TRUE),
      median_l = median(l, na.rm = TRUE),
      sd_l = sd(l, na.rm = TRUE),
      se_l = sd_l / sqrt(N - 1),
      mean_k = mean(k, na.rm = TRUE),
      median_k = median(k, na.rm = TRUE),
      sd_k = sd(k, na.rm = TRUE),
      se_k = sd_k / sqrt(N - 1),
      mean_r = mean(r, na.rm = TRUE),
      median_r = median(r, na.rm = TRUE),
      sd_r = sd(r, na.rm = TRUE),
      se_r = sd_r / sqrt(N - 1),
      mean_auc = mean(auc, na.rm = TRUE),
      median_auc = median(auc, na.rm = TRUE),
      sd_auc = sd(auc, na.rm = TRUE),
      se_auc = sd_auc / sqrt(N - 1),
      NG = sum(NG, na.rm = TRUE),
      DB =  sum(DB, na.rm = TRUE),
      SM = sum(SM, na.rm = TRUE)
    )

    # Get shift vals
    # if L is 0, that means the no growth on no drug
    # if L is NA at 0, that means the spot was removed due to contamination
    # if L is 0, keep the shift at 0 and for other drug concs calculate delta Ls with no shift
    # otherwise calculate shift at no drug conc
    if (is.na(x_stats_interaction$mean_l[1]) || x_stats_interaction$mean_l[1] == 0) {
      x_stats_interaction$Raw_Shift_l <- 0
      x_stats_interaction$Raw_Shift_k <- 0
      x_stats_interaction$Raw_Shift_r <- 0
      x_stats_interaction$Raw_Shift_auc <- 0
      x_stats_interaction$z_shift_l <- 0
      x_stats_interaction$z_shift_k <- 0
      x_stats_interaction$z_shift_r <- 0
      x_stats_interaction$z_shift_auc <- 0
    } else {
      x_stats_interaction$Raw_Shift_l <- x_stats_interaction$mean_l[1] - background_l
      x_stats_interaction$Raw_Shift_k <- x_stats_interaction$mean_k[1] - background_k
      x_stats_interaction$Raw_Shift_r <- x_stats_interaction$mean_r[1] - background_r
      x_stats_interaction$Raw_Shift_auc <- x_stats_interaction$mean_auc[1] - background_auc
      x_stats_interaction$z_shift_l <- x_stats_interaction$Raw_Shift_l[1] / x_stats_by_l$sd[1]
      x_stats_interaction$z_shift_k <- x_stats_interaction$Raw_Shift_k[1] / x_stats_by_k$sd[1]
      x_stats_interaction$z_shift_r <- x_stats_interaction$Raw_Shift_r[1] / x_stats_by_r$sd[1]
      x_stats_interaction$z_shift_auc <- x_stats_interaction$Raw_Shift_auc[1] / x_stats_by_auc$sd[1]
    }

    # Get WT vals
    x_stats_interaction$WT_l <- x_stats_by_l$mean
    x_stats_interaction$WT_k <- x_stats_by_k$mean
    x_stats_interaction$WT_r <- x_stats_by_r$mean
    x_stats_interaction$WT_auc <- x_stats_by_auc$mean

    # Get WT SD
    x_stats_interaction$WT_sd_l <- x_stats_by_l$sd
    x_stats_interaction$WT_sd_k <- x_stats_by_k$sd
    x_stats_interaction$WT_sd_r <- x_stats_by_r$sd
    x_stats_interaction$WT_sd_auc <- x_stats_by_auc$sd

    # Only get scores if there's growth at no drug
    if (x_stats_interaction$mean_l[1] != 0 && !is.na(x_stats_interaction$mean_l[1])) {

      # Calculate expected values
      x_stats_interaction$Exp_l <- x_stats_interaction$WT_l + x_stats_interaction$Raw_Shift_l
      x_stats_interaction$Exp_k <- x_stats_interaction$WT_k + x_stats_interaction$Raw_Shift_k
      x_stats_interaction$Exp_r <- x_stats_interaction$WT_r + x_stats_interaction$Raw_Shift_r
      x_stats_interaction$Exp_auc <- x_stats_interaction$WT_auc + x_stats_interaction$Raw_Shift_auc

      # Calculate normalized delta values
      x_stats_interaction$delta_l <- x_stats_interaction$mean_l - x_stats_interaction$Exp_l
      x_stats_interaction$delta_k <- x_stats_interaction$mean_k - x_stats_interaction$Exp_k
      x_stats_interaction$delta_r <- x_stats_interaction$mean_r - x_stats_interaction$Exp_r
      x_stats_interaction$delta_auc <- x_stats_interaction$mean_auc - x_stats_interaction$Exp_auc

      # Disregard shift for no growth values in Z score calculation
      if (sum(x_stats_interaction$NG, na.rm = TRUE) > 0) {
        x_stats_interaction[x_stats_interaction$NG == 1, ]$delta_l <-
          x_stats_interaction[x_stats_interaction$NG == 1, ]$mean_l - x_stats_interaction[x_stats_interaction$NG == 1, ]$WT_l
        x_stats_interaction[x_stats_interaction$NG == 1, ]$delta_k <-
          x_stats_interaction[x_stats_interaction$NG == 1, ]$mean_k - x_stats_interaction[x_stats_interaction$NG == 1, ]$WT_k
        x_stats_interaction[x_stats_interaction$NG == 1, ]$delta_r <-
          x_stats_interaction[x_stats_interaction$NG == 1, ]$mean_r - x_stats_interaction[x_stats_interaction$NG == 1, ]$WT_r
        x_stats_interaction[x_stats_interaction$NG == 1, ]$delta_auc <-
          x_stats_interaction[x_stats_interaction$NG == 1, ]$mean_auc - x_stats_interaction[x_stats_interaction$NG == 1, ]$WT_auc

      }
      # Disregard shift for set to max values in Z score calculation
      if (sum(x_stats_interaction$SM, na.rm = TRUE) > 0) {
        x_stats_interaction[x_stats_interaction$SM == 1, ]$delta_l <-
          x_stats_interaction[x_stats_interaction$SM == 1, ]$mean_l - x_stats_interaction[x_stats_interaction$SM == 1, ]$WT_l
        # Only calculate the L value without shift since L is the only adjusted value
        # x_stats_interaction[x_stats_interaction$SM == 1, ]$delta_k <-
        #   x_stats_interaction[x_stats_interaction$SM == 1, ]$mean_k - x_stats_interaction[x_stats_interaction$SM == 1, ]$WT_k
        # x_stats_interaction[x_stats_interaction$SM == 1, ]$delta_r <-
        #   x_stats_interaction[x_stats_interaction$SM == 1, ]$mean_r - x_stats_interaction[x_stats_interaction$SM == 1, ]$WT_r
        # x_stats_interaction[x_stats_interaction$SM == 1, ]$delta_auc <-
        #   x_stats_interaction[x_stats_interaction$SM == 1, ]$mean_auc - x_stats_interaction[x_stats_interaction$SM == 1, ]$WT_AU
      }

      # Calculate Z score at each concentration
      x_stats_interaction$zscore_l <- (x_stats_interaction$delta_l) / (x_stats_interaction$WT_sd_l)
      x_stats_interaction$zscore_k <- (x_stats_interaction$delta_k) / (x_stats_interaction$WT_sd_k)
      x_stats_interaction$zscore_r <- (x_stats_interaction$delta_r) / (x_stats_interaction$WT_sd_r)
      x_stats_interaction$zscore_auc <- (x_stats_interaction$delta_auc) / (x_stats_interaction$WT_sd_auc)

      # Get linear model
      gene_lm_l <- lm(formula = delta_l ~ conc_num_factor, data = x_stats_interaction)
      gene_lm_k <- lm(formula = delta_k ~ conc_num_factor, data = x_stats_interaction)
      gene_lm_r <- lm(formula = delta_r ~ conc_num_factor, data = x_stats_interaction)
      gene_lm_auc <- lm(formula = delta_auc ~ conc_num_factor, data = x_stats_interaction)

      # Get interaction score calculated by linear model and R-squared value for the fit
      gene_interaction_l <- max_conc * (gene_lm_l$coefficients[2]) + gene_lm_l$coefficients[1]
      r_squared_l <- summary(gene_lm_l)$r.squared
      gene_interaction_k <- max_conc * (gene_lm_k$coefficients[2]) + gene_lm_k$coefficients[1]
      r_squared_k <- summary(gene_lm_k)$r.squared
      gene_interaction_r <- max_conc * (gene_lm_r$coefficients[2]) + gene_lm_r$coefficients[1]
      r_squared_r <- summary(gene_lm_r)$r.squared
      gene_interaction_auc <- max_conc * (gene_lm_r$coefficients[2]) + gene_lm_auc$coefficients[1]
      r_squared_auc <- summary(gene_lm_auc)$r.squared

      # Get total of non removed values
      num_non_removed_conc <- total_conc_nums - sum(x_stats_interaction$DB, na.rm = TRUE) - 1

      # Report the scores
      interaction_scores_rf$OrfRep[interaction_scores_rf$OrfRep == gene_sel] <-
        x_gene_sel$OrfRep[1]
      interaction_scores_rf$Gene[interaction_scores_rf$OrfRep == gene_sel] <-
        x_gene_sel$Gene[1]
      interaction_scores_rf$Raw_Shift_l[interaction_scores_rf$OrfRep == gene_sel] <-
        x_stats_interaction$Raw_Shift_l[1]
      interaction_scores_rf$z_shift_l[interaction_scores_rf$OrfRep == gene_sel] <-
        x_stats_interaction$z_shift_l[1]
      interaction_scores_rf$lm_Score_l[interaction_scores_rf$OrfRep == gene_sel] <-
        gene_interaction_l
      interaction_scores_rf$R_Squared_l[interaction_scores_rf$OrfRep == gene_sel] <-
        r_squared_l
      interaction_scores_rf$Sum_z_Score_l[interaction_scores_rf$OrfRep == gene_sel] <-
        sum(x_stats_interaction$zscore_l, na.rm = TRUE)
      interaction_scores_rf$avg_zscore_l[interaction_scores_rf$OrfRep == gene_sel] <-
        sum(x_stats_interaction$zscore_l, na.rm = TRUE) / (num_non_removed_conc)
      interaction_scores_rf$Raw_Shift_k[interaction_scores_rf$OrfRep == gene_sel] <-
        x_stats_interaction$Raw_Shift_k[1]
      interaction_scores_rf$z_shift_k[interaction_scores_rf$OrfRep == gene_sel] <-
        x_stats_interaction$z_shift_k[1]
      interaction_scores_rf$lm_Score_k[interaction_scores_rf$OrfRep == gene_sel] <-
        gene_interaction_k
      interaction_scores_rf$R_Squared_k[interaction_scores_rf$OrfRep == gene_sel] <-
        r_squared_k
      interaction_scores_rf$Sum_z_Score_k[interaction_scores_rf$OrfRep == gene_sel] <-
        sum(x_stats_interaction$zscore_k, na.rm = TRUE)
      interaction_scores_rf$avg_zscore_k[interaction_scores_rf$OrfRep == gene_sel] <-
        sum(x_stats_interaction$zscore_k, na.rm = TRUE) / (num_non_removed_conc)
      interaction_scores_rf$Raw_Shift_r[interaction_scores_rf$OrfRep == gene_sel] <-
        x_stats_interaction$Raw_Shift_r[1]
      interaction_scores_rf$z_shift_r[interaction_scores_rf$OrfRep == gene_sel] <-
        x_stats_interaction$z_shift_r[1]
      interaction_scores_rf$lm_Score_r[interaction_scores_rf$OrfRep == gene_sel] <-
        gene_interaction_r
      interaction_scores_rf$R_Squared_r[interaction_scores_rf$OrfRep == gene_sel] <-
        r_squared_r
      interaction_scores_rf$Sum_z_Score_r[interaction_scores_rf$OrfRep == gene_sel] <-
        sum(x_stats_interaction$zscore_r, na.rm = TRUE)
      interaction_scores_rf$avg_zscore_r[interaction_scores_rf$OrfRep == gene_sel] <-
        sum(x_stats_interaction$zscore_r, na.rm = TRUE) / (total_conc_nums - 1)
      interaction_scores_rf$Raw_Shift_auc[interaction_scores_rf$OrfRep == gene_sel] <-
        x_stats_interaction$Raw_Shift_auc[1]
      interaction_scores_rf$z_shift_auc[interaction_scores_rf$OrfRep == gene_sel] <-
        x_stats_interaction$z_shift_auc[1]
      interaction_scores_rf$lm_Score_auc[interaction_scores_rf$OrfRep == gene_sel] <-
        gene_interaction_auc
      interaction_scores_rf$R_Squared_auc[interaction_scores_rf$OrfRep == gene_sel] <-
        r_squared_auc
      interaction_scores_rf$Sum_z_Score_auc[interaction_scores_rf$OrfRep == gene_sel] <-
        sum(x_stats_interaction$zscore_auc, na.rm = TRUE)
      interaction_scores_rf$avg_zscore_auc[interaction_scores_rf$OrfRep == gene_sel] <-
        sum(x_stats_interaction$zscore_auc, na.rm = TRUE) / (total_conc_nums - 1)
    }

    if (x_stats_interaction$mean_l[1] == 0 || is.na(x_stats_interaction$mean_l[1])) {

      # Calculate expected values
      x_stats_interaction$Exp_l <- x_stats_interaction$WT_l + x_stats_interaction$Raw_Shift_l
      x_stats_interaction$Exp_k <- x_stats_interaction$WT_k + x_stats_interaction$Raw_Shift_k
      x_stats_interaction$Exp_r <- x_stats_interaction$WT_r + x_stats_interaction$Raw_Shift_r
      x_stats_interaction$Exp_auc <- x_stats_interaction$WT_auc + x_stats_interaction$Raw_Shift_auc

      # Calculate normalized delta values
      x_stats_interaction$delta_l <- x_stats_interaction$mean_l - x_stats_interaction$Exp_l
      x_stats_interaction$delta_k <- x_stats_interaction$mean_k - x_stats_interaction$Exp_k
      x_stats_interaction$delta_r <- x_stats_interaction$mean_r - x_stats_interaction$Exp_r
      x_stats_interaction$delta_auc <- x_stats_interaction$mean_auc - x_stats_interaction$Exp_auc

      # Disregard shift for missing values in Z score calculation
      if (sum(x_stats_interaction$NG, na.rm = TRUE) > 0) {
        x_stats_interaction[x_stats_interaction$NG == 1, ]$delta_l <-
          x_stats_interaction[x_stats_interaction$NG == 1, ]$mean_l - x_stats_interaction[x_stats_interaction$NG == 1, ]$WT_l
        x_stats_interaction[x_stats_interaction$NG == 1, ]$delta_k <-
          x_stats_interaction[x_stats_interaction$NG == 1, ]$mean_k - x_stats_interaction[x_stats_interaction$NG == 1, ]$WT_k
        x_stats_interaction[x_stats_interaction$NG == 1, ]$delta_r <-
          x_stats_interaction[x_stats_interaction$NG == 1, ]$mean_r - x_stats_interaction[x_stats_interaction$NG == 1, ]$WT_r
        x_stats_interaction[x_stats_interaction$NG == 1, ]$delta_auc <-
          x_stats_interaction[x_stats_interaction$NG == 1, ]$mean_auc - x_stats_interaction[x_stats_interaction$NG == 1, ]$WT_auc
      }
      # Disregard shift for set to max values in Z score calculation
      if (sum(x_stats_interaction$SM, na.rm = TRUE) > 0) {
        x_stats_interaction[x_stats_interaction$SM == 1, ]$delta_l <-
          x_stats_interaction[x_stats_interaction$SM == 1, ]$mean_l - x_stats_interaction[x_stats_interaction$SM == 1, ]$WT_l
        # Only calculate the L value without shift since L is the only adjusted value
        # x_stats_interaction[x_stats_interaction$SM == 1, ]$delta_k <-
        #  x_stats_interaction[x_stats_interaction$SM == 1, ]$mean_k - x_stats_interaction[x_stats_interaction$SM == 1, ]$WT_k
        # x_stats_interaction[x_stats_interaction$SM == 1, ]$delta_r <-
        #   x_stats_interaction[x_stats_interaction$SM == 1, ]$mean_r - x_stats_interaction[x_stats_interaction$SM == 1, ]$WT_r
        # x_stats_interaction[x_stats_interaction$SM == 1, ]$delta_auc <-
        #   x_stats_interaction[x_stats_interaction$SM == 1, ]$mean_auc - x_stats_interaction[x_stats_interaction$SM == 1, ]$WT_auc
        
      }
      
      # Calculate Z score at each concentration
      x_stats_interaction$zscore_l <- (x_stats_interaction$delta_l) / (x_stats_interaction$WT_sd_l)
      x_stats_interaction$zscore_k <- (x_stats_interaction$delta_k) / (x_stats_interaction$WT_sd_k)
      x_stats_interaction$zscore_r <- (x_stats_interaction$delta_r) / (x_stats_interaction$WT_sd_r)
      x_stats_interaction$zscore_auc <- (x_stats_interaction$delta_auc) / (x_stats_interaction$WT_sd_auc)
      
      # NA values for the next part since there's an NA or 0 at the no drug.
      gene_lm_l <- NA
      gene_lm_k <- NA
      gene_lm_r <- NA
      gene_lm_auc <- NA
      gene_interaction_l <- NA
      r_squared_l <- NA
      gene_interaction_k <- NA
      r_squared_k <- NA
      gene_interaction_r <- NA
      r_squared_r <- NA
      gene_interaction_auc <- NA
      r_squared_auc <- NA
      x_stats_interaction$Raw_Shift_l <- NA
      x_stats_interaction$Raw_Shift_k <- NA
      x_stats_interaction$Raw_Shift_r <- NA
      x_stats_interaction$Raw_Shift_auc <- NA
      x_stats_interaction$z_shift_l <- NA
      x_stats_interaction$z_shift_k <- NA
      x_stats_interaction$z_shift_r <- NA
      x_stats_interaction$z_shift_auc <- NA
      interaction_scores_rf$OrfRep[interaction_scores_rf$OrfRep == gene_sel] <- x_gene_sel$OrfRep[1]
      interaction_scores_rf$Gene[interaction_scores_rf$OrfRep == gene_sel] <- x_gene_sel$Gene[1]
      interaction_scores_rf$Raw_Shift_l[interaction_scores_rf$OrfRep == gene_sel] <- x_stats_interaction$Raw_Shift_l[1]
      interaction_scores_rf$z_shift_l[interaction_scores_rf$OrfRep == gene_sel] <- x_stats_interaction$z_shift_l[1]
      interaction_scores_rf$lm_Score_l[interaction_scores_rf$OrfRep == gene_sel] <- gene_interaction_l
      interaction_scores_rf$R_Squared_l[interaction_scores_rf$OrfRep == gene_sel] <- r_squared_l
      interaction_scores_rf$Sum_z_Score_l[interaction_scores_rf$OrfRep == gene_sel] <- NA
      interaction_scores_rf$avg_zscore_l[interaction_scores_rf$OrfRep == gene_sel] <- NA
      interaction_scores_rf$Raw_Shift_k[interaction_scores_rf$OrfRep == gene_sel] <- x_stats_interaction$Raw_Shift_k[1]
      interaction_scores_rf$z_shift_k[interaction_scores_rf$OrfRep == gene_sel] <- x_stats_interaction$z_shift_k[1]
      interaction_scores_rf$lm_Score_k[interaction_scores_rf$OrfRep == gene_sel] <- gene_interaction_k
      interaction_scores_rf$R_Squared_k[interaction_scores_rf$OrfRep == gene_sel] <- r_squared_k
      interaction_scores_rf$Sum_z_Score_k[interaction_scores_rf$OrfRep == gene_sel] <- NA
      interaction_scores_rf$avg_zscore_k[interaction_scores_rf$OrfRep == gene_sel] <- NA
      interaction_scores_rf$Raw_Shift_r[interaction_scores_rf$OrfRep == gene_sel] <- x_stats_interaction$Raw_Shift_r[1]
      interaction_scores_rf$z_shift_r[interaction_scores_rf$OrfRep == gene_sel] <- x_stats_interaction$z_shift_r[1]
      interaction_scores_rf$lm_Score_r[interaction_scores_rf$OrfRep == gene_sel] <- gene_interaction_r
      interaction_scores_rf$R_Squared_r[interaction_scores_rf$OrfRep == gene_sel] <- r_squared_r
      interaction_scores_rf$Sum_z_Score_r[interaction_scores_rf$OrfRep == gene_sel] <- NA
      interaction_scores_rf$avg_zscore_r[interaction_scores_rf$OrfRep == gene_sel] <- NA
      interaction_scores_rf$Raw_Shift_auc[interaction_scores_rf$OrfRep == gene_sel] <- x_stats_interaction$Raw_Shift_auc[1]
      interaction_scores_rf$z_shift_auc[interaction_scores_rf$OrfRep == gene_sel] <- x_stats_interaction$z_shift_auc[1]
      interaction_scores_rf$lm_Score_auc[interaction_scores_rf$OrfRep == gene_sel] <- gene_interaction_auc
      interaction_scores_rf$R_Squared_auc[interaction_scores_rf$OrfRep == gene_sel] <- r_squared_auc
      interaction_scores_rf$Sum_z_Score_auc[interaction_scores_rf$OrfRep == gene_sel] <- NA
      interaction_scores_rf$avg_zscore_auc[interaction_scores_rf$OrfRep == gene_sel] <- NA
    }
    
    if (i == 1) {
      x_stats_interaction_all_rf <- x_stats_interaction
    }

    if (i > 1) {
      x_stats_interaction_all_rf <- rbind(x_stats_interaction_all_rf, x_stats_interaction)
    }
    
    interaction_scores_rf$NG[interaction_scores_rf$OrfRep == gene_sel] <- sum(x_stats_interaction$NG, na.rm = TRUE)
    interaction_scores_rf$DB[interaction_scores_rf$OrfRep == gene_sel] <- sum(x_stats_interaction$DB, na.rm = TRUE)
    interaction_scores_rf$SM[interaction_scores_rf$OrfRep == gene_sel] <- sum(x_stats_interaction$SM, na.rm = TRUE)
    
    # x_stats_l_int_temp <- rbind(x_stats_l_int_temp, x_stats_l_int)
  }

  print("Pass RF Calculation loop")
  
  lm_sd_l <- sd(interaction_scores_rf$lm_Score_l, na.rm = TRUE)
  lm_sd_k <- sd(interaction_scores_rf$lm_Score_k, na.rm = TRUE)
  lm_sd_r <- sd(interaction_scores_rf$lm_Score_r, na.rm = TRUE)
  lm_sd_auc <- sd(interaction_scores_rf$lm_Score_auc, na.rm = TRUE)
  lm_mean_l <- mean(interaction_scores_rf$lm_Score_l, na.rm = TRUE)
  lm_mean_k <- mean(interaction_scores_rf$lm_Score_k, na.rm = TRUE)
  lm_mean_r <- mean(interaction_scores_rf$lm_Score_r, na.rm = TRUE)
  lm_mean_auc <- mean(interaction_scores_rf$lm_Score_auc, na.rm = TRUE)
  
  print(paste("Mean RF linear regression score L", lm_mean_l))
  
  interaction_scores_rf$z_lm_l <- (interaction_scores_rf$lm_Score_l - lm_mean_l) / (lm_sd_l)
  interaction_scores_rf$z_lm_k <- (interaction_scores_rf$lm_Score_k - lm_mean_k) / (lm_sd_k)
  interaction_scores_rf$z_lm_r <- (interaction_scores_rf$lm_Score_r - lm_mean_r) / (lm_sd_r)
  interaction_scores_rf$z_lm_auc <- (interaction_scores_rf$lm_Score_auc - lm_mean_auc) / (lm_sd_auc)
  interaction_scores_rf <- interaction_scores_rf[order(interaction_scores_rf$z_lm_l, decreasing = TRUE), ]
  interaction_scores_rf <- interaction_scores_rf[order(interaction_scores_rf$NG, decreasing = TRUE), ]
  write.csv(interaction_scores_rf, file.path(out_dir, "rf_zscores_interaction.csv"), row.names = FALSE)
  
  for (i in 1:num_genes_rf) {
    gene_sel <- unique(interaction_scores_rf$OrfRep)[i]
    x_z_calculations <- x_stats_interaction_all_rf[x_stats_interaction_all_rf$OrfRep == gene_sel, ]
    df_int_scores <- interaction_scores_rf[interaction_scores_rf$OrfRep == gene_sel, ]
  
    p_rf_l[[i]] <- ggplot(x_z_calculations, aes(conc_num_factor, delta_l)) +
      geom_point() +
      geom_smooth(method = "lm", formula = y ~ x, se = FALSE) +
      coord_cartesian(ylim = c(-65, 65)) +
      geom_errorbar(aes(ymin = 0 - (2 * WT_sd_l), ymax = 0 + (2 * WT_sd_l)), alpha = 0.3) +
      ggtitle(paste(x_z_calculations$OrfRep[1], x_z_calculations$Gene[1], sep = "      ")) +
      annotate("text", x = 1, y = 45, label = paste("ZShift =", round(df_int_scores$z_shift_l, 2))) +
      scale_color_discrete(guide = FALSE) +
      # annotate("text", x = 1, y = 35, label = paste("Avg Zscore =", round(df_int_scores$avg_zscore_l, 2))) +
      annotate("text", x = 1, y = 25, label = paste("lm Zscore =", round(df_int_scores$z_lm_l, 2))) +
      annotate("text", x = 1, y = -25, label = paste("NG =", df_int_scores$NG)) +
      annotate("text", x = 1, y = -35, label = paste("DB =", df_int_scores$DB)) +
      annotate("text", x = 1, y = -45, label = paste("SM =", df_int_scores$SM)) +
      scale_x_continuous(
        name = unique(df$Drug[1]),
        breaks = unique(x_z_calculations$conc_num_factor),
        labels = unique(as.character(x_z_calculations$conc_num))
      ) +
      scale_y_continuous(breaks = c(-60, -50, -40, -30, -20, -10, 0, 10, 20, 30, 40, 50, 60)) +
      theme_publication()
    
    p_rf_k[[i]] <- ggplot(
      x_z_calculations, aes(conc_num_factor, delta_k)) +
      geom_point() +
      geom_smooth(method = "lm", formula = y ~ x, se = FALSE) +
      coord_cartesian(ylim = c(-65, 65)) +
      geom_errorbar(aes(ymin = 0 - (2 * WT_sd_k), ymax = 0 + (2 * WT_sd_k)), alpha = 0.3) +
      ggtitle(paste(x_z_calculations$OrfRep[1], x_z_calculations$Gene[1], sep = "      ")) +
      annotate("text", x = 1, y = 45, label = paste("ZShift =", round(df_int_scores$z_shift_k, 2))) +
      scale_color_discrete(guide = FALSE) +
      # annotate("text", x = 1, y = 35, label = paste("Avg ZScore =", round(df_int_scores$avg_zscore_k, 2))) +
      annotate("text", x = 1, y = 25, label = paste("lm ZScore =", round(df_int_scores$z_lm_l, 2))) +
      annotate("text", x = 1, y = -25, label = paste("NG =", df_int_scores$NG)) +
      annotate("text", x = 1, y = -35, label = paste("DB =", df_int_scores$DB)) +
      annotate("text", x = 1, y = -45, label = paste("SM =", df_int_scores$SM)) +
      scale_x_continuous(
        name = unique(df$Drug[1]),
        breaks = unique(x_z_calculations$conc_num_factor),
        labels = unique(as.character(x_z_calculations$conc_num))
      ) +
      scale_y_continuous(breaks = c(-60, -50, -40, -30, -20, -10, 0, 10, 20, 30, 40, 50, 60)) +
      theme_publication()
    
    p_rf_r[[i]] <- ggplot(
      x_z_calculations, aes(conc_num_factor, delta_r)) +
      geom_point() +
      geom_smooth(method = "lm", formula = y ~ x, se = FALSE) +
      coord_cartesian(ylim = c(-0.65, 0.65)) +
      geom_errorbar(aes(ymin = 0 - (2 * WT_sd_r), ymax = 0 + (2 * WT_sd_r)), alpha = 0.3) +
      ggtitle(paste(x_z_calculations$OrfRep[1], x_z_calculations$Gene[1], sep = "      ")) +
      annotate("text", x = 1, y = 0.45, label = paste("ZShift =", round(df_int_scores$z_shift_r, 2))) +
      scale_color_discrete(guide = FALSE) +
      # annotate("text", x = 1, y = 0.35, label = paste("Avg ZScore =", round(df_int_scores$avg_zscore_r, 2))) +
      annotate("text", x = 1, y = 0.25, label = paste("lm ZScore =", round(df_int_scores$z_lm_r, 2))) +
      annotate("text", x = 1, y = -0.25, label = paste("NG =", df_int_scores$NG)) +
      annotate("text", x = 1, y = -0.35, label = paste("DB =", df_int_scores$DB)) +
      annotate("text", x = 1, y = -0.45, label = paste("SM =", df_int_scores$SM)) +
      scale_x_continuous(
        name = unique(df$Drug[1]),
        breaks = unique(x_z_calculations$conc_num_factor),
        labels = unique(as.character(x_z_calculations$conc_num))
      ) +
      scale_y_continuous(breaks = c(-0.6, -0.4, -0.2, 0, 0.2, 0.4, 0.6)) +
      theme_publication()
   
    p_rf_auc[[i]] <- ggplot(
      x_z_calculations, aes(conc_num_factor, delta_auc)) +
      geom_point() +
      geom_smooth(method = "lm", formula = y ~ x, se = FALSE) +
      coord_cartesian(ylim = c(-6500, 6500)) +
      geom_errorbar(aes(ymin = 0 - (2 * WT_sd_auc), ymax = 0 + (2 * WT_sd_auc)), alpha = 0.3) +
      ggtitle(paste(x_z_calculations$OrfRep[1], x_z_calculations$Gene[1], sep = "      ")) +
      annotate("text", x = 1, y = 4500, label = paste("ZShift =", round(df_int_scores$z_shift_auc, 2))) +
      scale_color_discrete(guide = FALSE) +
      # annotate("text", x = 1, y = 3500, label = paste("Avg ZScore =", round(df_int_scores$avg_zscore_auc, 2))) +
      annotate("text", x = 1, y = 2500, label = paste("lm ZScore =", round(df_int_scores$z_lm_auc, 2))) +
      annotate("text", x = 1, y = -2500, label = paste("NG =", df_int_scores$NG)) +
      annotate("text", x = 1, y = -3500, label = paste("DB =", df_int_scores$DB)) +
      annotate("text", x = 1, y = -4500, label = paste("SM =", df_int_scores$SM)) +
      scale_x_continuous(
        name = unique(df$Drug[1]),
        breaks = unique(x_z_calculations$conc_num_factor),
        labels = unique(as.character(x_z_calculations$conc_num))
      ) +
      scale_y_continuous(breaks = c(-6000, -5000, -4000, -3000, -2000, -1000, 0, 1000, 2000, 3000, 4000, 5000, 6000)) +
      theme_publication()
    
    if (i == 1) {
      x_stats_interaction_all_rf_final <- x_z_calculations
    }
    if (i > 1) {
      x_stats_interaction_all_rf_final <- rbind(x_stats_interaction_all_rf_final, x_z_calculations)
    }
  }
  print("Pass RF ggplot loop")
  write.csv(x_stats_interaction_all_rf_final, file.path(out_dir, "rf_zscore_calculations.csv"), row.names = FALSE)
  
  # Part 5 - Get Zscores for Gene deletion strains
  
  # Get total number of genes for the next loop
  num_genes <- length(unique(df2$OrfRep))
  # print(num_genes)
  
  # Create the output data.frame containing columns for each deletion strain
  interaction_scores <- unique(df2["OrfRep"])
  # interaction_scores$Gene <- unique(df2$Gene)
  interaction_scores$Gene <- NA
  interaction_scores$Raw_Shift_l <- NA
  interaction_scores$z_shift_l <- NA
  interaction_scores$lm_Score_l <- NA
  interaction_scores$z_lm_l <- NA
  interaction_scores$R_Squared_l <- NA
  interaction_scores$Sum_z_Score_l <- NA
  interaction_scores$avg_zscore_l <- NA
  interaction_scores$Raw_Shift_k <- NA
  interaction_scores$z_shift_k <- NA
  interaction_scores$lm_Score_k <- NA
  interaction_scores$z_lm_k <- NA
  interaction_scores$R_Squared_k <- NA
  interaction_scores$Sum_z_Score_k <- NA
  interaction_scores$avg_zscore_k <- NA
  interaction_scores$Raw_Shift_r <- NA
  interaction_scores$z_shift_r <- NA
  interaction_scores$lm_Score_r <- NA
  interaction_scores$z_lm_r <- NA
  interaction_scores$R_Squared_r <- NA
  interaction_scores$Sum_z_Score_r <- NA
  interaction_scores$avg_zscore_r <- NA
  interaction_scores$Raw_Shift_auc <- NA
  interaction_scores$z_shift_auc <- NA
  interaction_scores$lm_Score_auc <- NA
  interaction_scores$z_lm_auc <- NA
  interaction_scores$R_Squared_auc <- NA
  interaction_scores$Sum_z_Score_auc <- NA
  interaction_scores$avg_zscore_auc <- NA
  interaction_scores$NG <- NA
  interaction_scores$DB <- NA
  interaction_scores$SM <- NA
  
  for (i in 1:num_genes) {
    # Get each deletion strain ORF
    gene_sel <- unique(df2$OrfRep)[i]
    # Extract only the current deletion strain and its data
    x_gene_sel <- df2[df2$OrfRep == gene_sel, ]
    
    x_stats_interaction <- ddply(
      x_gene_sel,
      c("OrfRep", "Gene", "conc_num", "conc_num_factor"),
      summarise,
      N = (length(l)),
      mean_l = mean(l, na.rm = TRUE),
      median_l = median(l, na.rm = TRUE),
      sd_l = sd(l, na.rm = TRUE),
      se_l = sd_l / sqrt(N - 1),
      mean_k = mean(k, na.rm = TRUE),
      median_k = median(k, na.rm = TRUE),
      sd_k = sd(k, na.rm = TRUE),
      se_k = sd_k / sqrt(N - 1),
      mean_r = mean(r, na.rm = TRUE),
      median_r = median(r, na.rm = TRUE),
      sd_r = sd(r, na.rm = TRUE),
      se_r = sd_r / sqrt(N - 1),
      mean_auc = mean(auc, na.rm = TRUE),
      median_auc = median(auc, na.rm = TRUE),
      sd_auc = sd(auc, na.rm = TRUE),
      se_auc = sd_auc / sqrt(N - 1),
      NG = sum(NG, na.rm = TRUE),
      DB = sum(DB, na.rm = TRUE),
      SM = sum(SM, na.rm = TRUE)
    )
    
    # Get shift vals
    # if L is 0, that means the no growth on no drug
    # if L is NA at 0, that means the spot was removed due to contamination
    # if L is 0, keep the shift at 0 and for other drug concs calculate delta Ls with no shift
    # otherwise calculate shift at no drug conc
    if (is.na(x_stats_interaction$mean_l[1]) || x_stats_interaction$mean_l[1] == 0) {
      x_stats_interaction$Raw_Shift_l <- 0
      x_stats_interaction$Raw_Shift_k <- 0
      x_stats_interaction$Raw_Shift_r <- 0
      x_stats_interaction$Raw_Shift_auc <- 0
      x_stats_interaction$z_shift_l <- 0
      x_stats_interaction$z_shift_k <- 0
      x_stats_interaction$z_shift_r <- 0
      x_stats_interaction$z_shift_auc <- 0
    } else {
      x_stats_interaction$Raw_Shift_l <- x_stats_interaction$mean_l[1] - background_l
      x_stats_interaction$Raw_Shift_k <- x_stats_interaction$mean_k[1] - background_k
      x_stats_interaction$Raw_Shift_r <- x_stats_interaction$mean_r[1] - background_r
      x_stats_interaction$Raw_Shift_auc <- x_stats_interaction$mean_auc[1] - background_auc
      x_stats_interaction$z_shift_l <- x_stats_interaction$Raw_Shift_l[1] / x_stats_by_l$sd[1]
      x_stats_interaction$z_shift_k <- x_stats_interaction$Raw_Shift_k[1] / x_stats_by_k$sd[1]
      x_stats_interaction$z_shift_r <- x_stats_interaction$Raw_Shift_r[1] / x_stats_by_r$sd[1]
      x_stats_interaction$z_shift_auc <- x_stats_interaction$Raw_Shift_auc[1] / x_stats_by_auc$sd[1]
    }
    
    # Get WT vals
    x_stats_interaction$WT_l <- x_stats_by_l$mean
    x_stats_interaction$WT_k <- x_stats_by_k$mean
    x_stats_interaction$WT_r <- x_stats_by_r$mean
    x_stats_interaction$WT_auc <- x_stats_by_auc$mean
    
    # Get WT SD
    x_stats_interaction$WT_sd_l <- x_stats_by_l$sd
    x_stats_interaction$WT_sd_k <- x_stats_by_k$sd
    x_stats_interaction$WT_sd_r <- x_stats_by_r$sd
    x_stats_interaction$WT_sd_auc <- x_stats_by_auc$sd
    
    # Only get scores if there's growth at no drug
    if (x_stats_interaction$mean_l[1] != 0 && !is.na(x_stats_interaction$mean_l[1])) {
      
      # Calculate expected values
      x_stats_interaction$Exp_l <- x_stats_interaction$WT_l + x_stats_interaction$Raw_Shift_l
      x_stats_interaction$Exp_k <- x_stats_interaction$WT_k + x_stats_interaction$Raw_Shift_k
      x_stats_interaction$Exp_r <- x_stats_interaction$WT_r + x_stats_interaction$Raw_Shift_r
      x_stats_interaction$Exp_auc <- x_stats_interaction$WT_auc + x_stats_interaction$Raw_Shift_auc
      
      # Calculate normalized delta values
      x_stats_interaction$delta_l <- x_stats_interaction$mean_l - x_stats_interaction$Exp_l
      x_stats_interaction$delta_k <- x_stats_interaction$mean_k - x_stats_interaction$Exp_k
      x_stats_interaction$delta_r <- x_stats_interaction$mean_r - x_stats_interaction$Exp_r
      x_stats_interaction$delta_auc <- x_stats_interaction$mean_auc - x_stats_interaction$Exp_auc
      
      # Disregard shift for no growth values in Z score calculation
      if (sum(x_stats_interaction$NG, na.rm = TRUE) > 0) {
        x_stats_interaction[x_stats_interaction$NG == 1, ]$delta_l <-
          x_stats_interaction[x_stats_interaction$NG == 1, ]$mean_l - x_stats_interaction[x_stats_interaction$NG == 1, ]$WT_l
        x_stats_interaction[x_stats_interaction$NG == 1, ]$delta_k <-
          x_stats_interaction[x_stats_interaction$NG == 1, ]$mean_k - x_stats_interaction[x_stats_interaction$NG == 1, ]$WT_k
        x_stats_interaction[x_stats_interaction$NG == 1, ]$delta_r <-
          x_stats_interaction[x_stats_interaction$NG == 1, ]$mean_r - x_stats_interaction[x_stats_interaction$NG == 1, ]$WT_r
        x_stats_interaction[x_stats_interaction$NG == 1, ]$delta_auc <-
          x_stats_interaction[x_stats_interaction$NG == 1, ]$mean_auc - x_stats_interaction[x_stats_interaction$NG == 1, ]$WT_auc
      }

      # Disregard shift for set to max values in Z score calculation
      if (sum(x_stats_interaction$SM, na.rm = TRUE) > 0) {
        x_stats_interaction[x_stats_interaction$SM == 1, ]$delta_l <-
          x_stats_interaction[x_stats_interaction$SM == 1, ]$mean_l - x_stats_interaction[x_stats_interaction$SM == 1, ]$WT_l
        # Only calculate the L value without shift since L is the only adjusted value
        # x_stats_interaction[x_stats_interaction$SM == 1, ]$delta_k <-
        #   x_stats_interaction[x_stats_interaction$SM == 1, ]$mean_k - x_stats_interaction[x_stats_interaction$SM == 1, ]$WT_k
        # x_stats_interaction[x_stats_interaction$SM == 1, ]$delta_r <-
        #   x_stats_interaction[x_stats_interaction$SM == 1, ]$mean_r - x_stats_interaction[x_stats_interaction$SM == 1, ]$WT_r
        # x_stats_interaction[x_stats_interaction$SM == 1, ]$delta_auc <-
        #   x_stats_interaction[x_stats_interaction$SM == 1, ]$mean_auc - x_stats_interaction[x_stats_interaction$SM == 1, ]$WT_auc
      }
      
      # Calculate Z score at each concentration
      x_stats_interaction$zscore_l <- (x_stats_interaction$delta_l) / (x_stats_interaction$WT_sd_l)
      x_stats_interaction$zscore_k <- (x_stats_interaction$delta_k) / (x_stats_interaction$WT_sd_k)
      x_stats_interaction$zscore_r <- (x_stats_interaction$delta_r) / (x_stats_interaction$WT_sd_r)
      x_stats_interaction$zscore_auc <- (x_stats_interaction$delta_auc) / (x_stats_interaction$WT_sd_auc)
      
      # Get linear model
      gene_lm_l <- lm(formula = delta_l ~ conc_num_factor, data = x_stats_interaction)
      gene_lm_k <- lm(formula = delta_k ~ conc_num_factor, data = x_stats_interaction)
      gene_lm_r <- lm(formula = delta_r ~ conc_num_factor, data = x_stats_interaction)
      gene_lm_auc <- lm(formula = delta_auc ~ conc_num_factor, data = x_stats_interaction)
      
      # Get interaction score calculated by linear model and R-squared value for the fit
      gene_interaction_l <- max_conc * (gene_lm_l$coefficients[2]) + gene_lm_l$coefficients[1]
      r_squared_l <- summary(gene_lm_l)$r.squared
      gene_interaction_k <- max_conc * (gene_lm_k$coefficients[2]) + gene_lm_k$coefficients[1]
      r_squared_k <- summary(gene_lm_k)$r.squared
      gene_interaction_r <- max_conc * (gene_lm_r$coefficients[2]) + gene_lm_r$coefficients[1]
      r_squared_r <- summary(gene_lm_r)$r.squared
      gene_interaction_auc <- max_conc * (gene_lm_r$coefficients[2]) + gene_lm_auc$coefficients[1]
      r_squared_auc <- summary(gene_lm_auc)$r.squared
      
      # Get total of non removed values
      num_non_removed_conc <- total_conc_nums - sum(x_stats_interaction$DB, na.rm = TRUE) - 1
      
      # Report the scores
      interaction_scores$OrfRep[interaction_scores$OrfRep == gene_sel] <-
        as.character(x_gene_sel$OrfRep[1])
      interaction_scores$Gene[interaction_scores$OrfRep == gene_sel] <-
        as.character(x_gene_sel$Gene[1])
      interaction_scores$Raw_Shift_l[interaction_scores$OrfRep == gene_sel] <-
        x_stats_interaction$Raw_Shift_l[1]
      interaction_scores$z_shift_l[interaction_scores$OrfRep == gene_sel] <-
        x_stats_interaction$z_shift_l[1]
      interaction_scores$lm_Score_l[interaction_scores$OrfRep == gene_sel] <-
        gene_interaction_l
      interaction_scores$z_lm_l[interaction_scores$OrfRep == gene_sel] <-
        (gene_interaction_l - lm_mean_l) / lm_sd_l
      interaction_scores$R_Squared_l[interaction_scores$OrfRep == gene_sel] <-
        r_squared_l
      interaction_scores$Sum_z_Score_l[interaction_scores$OrfRep == gene_sel] <-
        sum(x_stats_interaction$zscore_l, na.rm = TRUE)
      interaction_scores$avg_zscore_l[interaction_scores$OrfRep == gene_sel] <-
        sum(x_stats_interaction$zscore_l, na.rm = TRUE) / (num_non_removed_conc)
      interaction_scores$Raw_Shift_k[interaction_scores$OrfRep == gene_sel] <-
        x_stats_interaction$Raw_Shift_k[1]
      interaction_scores$z_shift_k[interaction_scores$OrfRep == gene_sel] <-
        x_stats_interaction$z_shift_k[1]
      interaction_scores$lm_Score_k[interaction_scores$OrfRep == gene_sel] <-
        gene_interaction_k
      interaction_scores$z_lm_k[interaction_scores$OrfRep == gene_sel] <-
        (gene_interaction_k - lm_mean_k) / lm_sd_k
      interaction_scores$R_Squared_k[interaction_scores$OrfRep == gene_sel] <-
        r_squared_k
      interaction_scores$Sum_z_Score_k[interaction_scores$OrfRep == gene_sel] <-
        sum(x_stats_interaction$zscore_k, na.rm = TRUE)
      interaction_scores$avg_zscore_k[interaction_scores$OrfRep == gene_sel] <-
        sum(x_stats_interaction$zscore_k, na.rm = TRUE) / (num_non_removed_conc)
      interaction_scores$Raw_Shift_r[interaction_scores$OrfRep == gene_sel] <-
        x_stats_interaction$Raw_Shift_r[1]
      interaction_scores$z_shift_r[interaction_scores$OrfRep == gene_sel] <-
        x_stats_interaction$z_shift_r[1]
      interaction_scores$lm_Score_r[interaction_scores$OrfRep == gene_sel] <-
        gene_interaction_r
      interaction_scores$z_lm_r[interaction_scores$OrfRep == gene_sel] <-
        (gene_interaction_r - lm_mean_r) / lm_sd_r
      interaction_scores$R_Squared_r[interaction_scores$OrfRep == gene_sel] <-
        r_squared_r
      interaction_scores$Sum_z_Score_r[interaction_scores$OrfRep == gene_sel] <-
        sum(x_stats_interaction$zscore_r, na.rm = TRUE)
      interaction_scores$avg_zscore_r[interaction_scores$OrfRep == gene_sel] <-
        sum(x_stats_interaction$zscore_r, na.rm = TRUE) / (total_conc_nums - 1)
      interaction_scores$Raw_Shift_auc[interaction_scores$OrfRep == gene_sel] <-
        x_stats_interaction$Raw_Shift_auc[1]
      interaction_scores$z_shift_auc[interaction_scores$OrfRep == gene_sel] <-
        x_stats_interaction$z_shift_auc[1]
      interaction_scores$lm_Score_auc[interaction_scores$OrfRep == gene_sel] <-
        gene_interaction_auc
      interaction_scores$z_lm_auc[interaction_scores$OrfRep == gene_sel] <-
        (gene_interaction_auc - lm_mean_auc) / lm_sd_auc
      interaction_scores$R_Squared_auc[interaction_scores$OrfRep == gene_sel] <-
        r_squared_auc
      interaction_scores$Sum_z_Score_auc[interaction_scores$OrfRep == gene_sel] <-
        sum(x_stats_interaction$zscore_auc, na.rm = TRUE)
      interaction_scores$avg_zscore_auc[interaction_scores$OrfRep == gene_sel] <-
        sum(x_stats_interaction$zscore_auc, na.rm = TRUE) / (total_conc_nums - 1)
    }
    
    if (x_stats_interaction$mean_l[1] == 0 || is.na(x_stats_interaction$mean_l[1])) {
      
      # Calculate expected values
      x_stats_interaction$Exp_l <- x_stats_interaction$WT_l + x_stats_interaction$Raw_Shift_l
      x_stats_interaction$Exp_k <- x_stats_interaction$WT_k + x_stats_interaction$Raw_Shift_k
      x_stats_interaction$Exp_r <- x_stats_interaction$WT_r + x_stats_interaction$Raw_Shift_r
      x_stats_interaction$Exp_auc <- x_stats_interaction$WT_auc + x_stats_interaction$Raw_Shift_auc
      
      # Calculate normalized delta values
      x_stats_interaction$delta_l <- x_stats_interaction$mean_l - x_stats_interaction$Exp_l
      x_stats_interaction$delta_k <- x_stats_interaction$mean_k - x_stats_interaction$Exp_k
      x_stats_interaction$delta_r <- x_stats_interaction$mean_r - x_stats_interaction$Exp_r
      x_stats_interaction$delta_auc <- x_stats_interaction$mean_auc - x_stats_interaction$Exp_auc
      
      # Disregard shift for missing values in Z score calculatiom
      if (sum(x_stats_interaction$NG, na.rm = TRUE) > 0) {
        x_stats_interaction[x_stats_interaction$NG == 1, ]$delta_l <-
          x_stats_interaction[x_stats_interaction$NG == 1, ]$mean_l - x_stats_interaction[x_stats_interaction$NG == 1, ]$WT_l
        x_stats_interaction[x_stats_interaction$NG == 1, ]$delta_k <-
          x_stats_interaction[x_stats_interaction$NG == 1, ]$mean_k - x_stats_interaction[x_stats_interaction$NG == 1, ]$WT_k
        x_stats_interaction[x_stats_interaction$NG == 1, ]$delta_r <-
          x_stats_interaction[x_stats_interaction$NG == 1, ]$mean_r - x_stats_interaction[x_stats_interaction$NG == 1, ]$WT_r
        x_stats_interaction[x_stats_interaction$NG == 1, ]$delta_auc <-
          x_stats_interaction[x_stats_interaction$NG == 1, ]$mean_auc - x_stats_interaction[x_stats_interaction$NG == 1, ]$WT_auc
      }

      # Disregard shift for set to max values in Z score calculation
      if (sum(x_stats_interaction$SM, na.rm = TRUE) > 0) {
        x_stats_interaction[x_stats_interaction$SM == 1, ]$delta_l <-
          x_stats_interaction[x_stats_interaction$SM == 1, ]$mean_l - x_stats_interaction[x_stats_interaction$SM == 1, ]$WT_l
        
        # Only calculate the L value without shift since L is the only adjusted value
        # x_stats_interaction[x_stats_interaction$SM == 1, ]$delta_k <-
        #   x_stats_interaction[x_stats_interaction$SM == 1, ]$mean_k - x_stats_interaction[x_stats_interaction$SM == 1, ]$WT_k
        # x_stats_interaction[x_stats_interaction$SM == 1, ]$delta_r <-
        #   x_stats_interaction[x_stats_interaction$SM == 1, ]$mean_r - x_stats_interaction[x_stats_interaction$SM == 1, ]$WT_r
        # x_stats_interaction[x_stats_interaction$SM == 1, ]$delta_auc <-
        #   x_stats_interaction[x_stats_interaction$SM == 1, ]$mean_auc - x_stats_interaction[x_stats_interaction$SM == 1, ]$WT_auc
      }
      
      # Calculate Z score at each concentration
      x_stats_interaction$zscore_l <- (x_stats_interaction$delta_l) / (x_stats_interaction$WT_sd_l)
      x_stats_interaction$zscore_k <- (x_stats_interaction$delta_k) / (x_stats_interaction$WT_sd_k)
      x_stats_interaction$zscore_r <- (x_stats_interaction$delta_r) / (x_stats_interaction$WT_sd_r)
      x_stats_interaction$zscore_auc <- (x_stats_interaction$delta_auc) / (x_stats_interaction$WT_sd_auc)
      
      # NA values for the next part since there's an NA or 0 at the no drug.
      gene_lm_l <- NA
      gene_lm_k <- NA
      gene_lm_r <- NA
      gene_interaction_l <- NA
      r_squared_l <- NA
      gene_interaction_k <- NA
      r_squared_k <- NA
      gene_interaction_r <- NA
      r_squared_r <- NA
      x_stats_interaction$Raw_Shift_l <- NA
      x_stats_interaction$Raw_Shift_k <- NA
      x_stats_interaction$Raw_Shift_r <- NA
      x_stats_interaction$Raw_Shift_auc <- NA
      x_stats_interaction$z_shift_l <- NA
      x_stats_interaction$z_shift_k <- NA
      x_stats_interaction$z_shift_r <- NA
      x_stats_interaction$z_shift_auc <- NA
      interaction_scores$OrfRep[interaction_scores$OrfRep == gene_sel] <- as.character(x_gene_sel$OrfRep[1])
      interaction_scores$Gene[interaction_scores$OrfRep == gene_sel] <- as.character(x_gene_sel$Gene[1])
      interaction_scores$Raw_Shift_l[interaction_scores$OrfRep == gene_sel] <- x_stats_interaction$Raw_Shift_l[1]
      interaction_scores$z_shift_l[interaction_scores$OrfRep == gene_sel] <- x_stats_interaction$z_shift_l[1]
      interaction_scores$lm_Score_l[interaction_scores$OrfRep == gene_sel] <- NA
      interaction_scores$z_lm_l[interaction_scores$OrfRep == gene_sel] <- NA
      interaction_scores$R_Squared_l[interaction_scores$OrfRep == gene_sel] <- r_squared_l
      interaction_scores$Sum_z_Score_l[interaction_scores$OrfRep == gene_sel] <- NA
      interaction_scores$avg_zscore_l[interaction_scores$OrfRep == gene_sel] <- NA
      interaction_scores$Raw_Shift_k[interaction_scores$OrfRep == gene_sel] <- x_stats_interaction$Raw_Shift_k[1]
      interaction_scores$z_shift_k[interaction_scores$OrfRep == gene_sel] <- x_stats_interaction$z_shift_k[1]
      interaction_scores$lm_Score_k[interaction_scores$OrfRep == gene_sel] <- NA
      interaction_scores$z_lm_k[interaction_scores$OrfRep == gene_sel] <- NA
      interaction_scores$R_Squared_k[interaction_scores$OrfRep == gene_sel] <- r_squared_k
      interaction_scores$Sum_z_Score_k[interaction_scores$OrfRep == gene_sel] <- NA
      interaction_scores$avg_zscore_k[interaction_scores$OrfRep == gene_sel] <- NA
      interaction_scores$Raw_Shift_r[interaction_scores$OrfRep == gene_sel] <- x_stats_interaction$Raw_Shift_r[1]
      interaction_scores$z_shift_r[interaction_scores$OrfRep == gene_sel] <- x_stats_interaction$z_shift_r[1]
      interaction_scores$lm_Score_r[interaction_scores$OrfRep == gene_sel] <- NA
      interaction_scores$z_lm_r[interaction_scores$OrfRep == gene_sel] <- NA
      interaction_scores$R_Squared_r[interaction_scores$OrfRep == gene_sel] <- r_squared_r
      interaction_scores$Sum_z_Score_r[interaction_scores$OrfRep == gene_sel] <- NA
      interaction_scores$avg_zscore_r[interaction_scores$OrfRep == gene_sel] <- NA
      interaction_scores$Raw_Shift_auc[interaction_scores$OrfRep == gene_sel] <- x_stats_interaction$Raw_Shift_auc[1]
      interaction_scores$z_shift_auc[interaction_scores$OrfRep == gene_sel] <- x_stats_interaction$z_shift_auc[1]
      interaction_scores$lm_Score_auc[interaction_scores$OrfRep == gene_sel] <- NA
      interaction_scores$z_lm_auc[interaction_scores$OrfRep == gene_sel] <- NA
      interaction_scores$R_Squared_auc[interaction_scores$OrfRep == gene_sel] <- r_squared_auc
      interaction_scores$Sum_z_Score_auc[interaction_scores$OrfRep == gene_sel] <- NA
      interaction_scores$avg_zscore_auc[interaction_scores$OrfRep == gene_sel] <- NA
    }
    
    if (i == 1) {
      x_stats_interaction_all <- x_stats_interaction
    }
    if (i > 1) {
      x_stats_interaction_all <- rbind(x_stats_interaction_all, x_stats_interaction)
    }
    
    interaction_scores$NG[interaction_scores$OrfRep == gene_sel] <- sum(x_stats_interaction$NG, na.rm = TRUE)
    interaction_scores$DB[interaction_scores$OrfRep == gene_sel] <- sum(x_stats_interaction$DB, na.rm = TRUE)
    interaction_scores$SM[interaction_scores$OrfRep == gene_sel] <- sum(x_stats_interaction$SM, na.rm = TRUE)
    
    # x_stats_l_int_temp <- rbind(x_stats_l_int_temp, x_stats_l_int)
  }

  print("Pass Int Calculation loop")
  interaction_scores <- interaction_scores[order(interaction_scores$z_lm_l, decreasing = TRUE), ]
  interaction_scores <- interaction_scores[order(interaction_scores$NG, decreasing = TRUE), ]
  df_order_by_OrfRep <- unique(interaction_scores$OrfRep)
  # x_stats_interaction_all <- x_stats_interaction_all[order(x_stats_interaction_all$NG, decreasing = TRUE), ]
  write.csv(interaction_scores, file.path(out_dir, "zscores_interaction.csv"), row.names = FALSE)
  
  interaction_scores_deletion_enhancers_l <-
    interaction_scores[interaction_scores$avg_zscore_l >= 2, ]
  interaction_scores_deletion_enhancers_k <-
    interaction_scores[interaction_scores$avg_zscore_k <= -2, ]
  interaction_scores_deletion_suppressors_l <-
    interaction_scores[interaction_scores$avg_zscore_l <= -2, ]
  interaction_scores_deletion_suppressors_k <-
    interaction_scores[interaction_scores$avg_zscore_k >= 2, ]
  interaction_scores_deletion_enhancers_and_suppressors_l <-
    interaction_scores[interaction_scores$avg_zscore_l >= 2 | interaction_scores$avg_zscore_l <= -2, ]
  interaction_scores_deletion_enhancers_and_suppressors_k <-
    interaction_scores[interaction_scores$avg_zscore_k >= 2 | interaction_scores$avg_zscore_k <= -2, ]
  interaction_scores_deletion_enhancers_lm_suppressors_avg_zscore_l <-
    interaction_scores[interaction_scores$z_lm_l >= 2 & interaction_scores$avg_zscore_l <= -2, ]
  interaction_scores_deletion_enhancers_avg_zscore_suppressors_lm_l <-
    interaction_scores[interaction_scores$z_lm_l <= -2 & interaction_scores$avg_zscore_l >= 2, ]
  interaction_scores_deletion_enhancers_lm_suppressors_avg_zscore_k <-
    interaction_scores[interaction_scores$z_lm_k <= -2 & interaction_scores$avg_zscore_k >= 2, ]
  interaction_scores_deletion_enhancers_avg_zscore_suppressors_lm_k <-
    interaction_scores[interaction_scores$z_lm_k >= 2 & interaction_scores$avg_zscore_k <= -2, ]
  interaction_scores_deletion_enhancers_l <-
    interaction_scores_deletion_enhancers_l[
      !is.na(interaction_scores_deletion_enhancers_l$OrfRep), ]
  interaction_scores_deletion_enhancers_k <-
    interaction_scores_deletion_enhancers_k[
      !is.na(interaction_scores_deletion_enhancers_k$OrfRep), ]
  interaction_scores_deletion_suppressors_l <-
    interaction_scores_deletion_suppressors_l[
      !is.na(interaction_scores_deletion_suppressors_l$OrfRep), ]
  interaction_scores_deletion_suppressors_k <-
    interaction_scores_deletion_suppressors_k[
      !is.na(interaction_scores_deletion_suppressors_k$OrfRep), ]
  interaction_scores_deletion_enhancers_and_suppressors_l <-
    interaction_scores_deletion_enhancers_and_suppressors_l[
      !is.na(interaction_scores_deletion_enhancers_and_suppressors_l$OrfRep), ]
  interaction_scores_deletion_enhancers_and_suppressors_k <-
    interaction_scores_deletion_enhancers_and_suppressors_k[
      !is.na(interaction_scores_deletion_enhancers_and_suppressors_k$OrfRep), ]
  interaction_scores_deletion_enhancers_lm_suppressors_avg_zscore_l <-
    interaction_scores_deletion_enhancers_lm_suppressors_avg_zscore_l[
      !is.na(interaction_scores_deletion_enhancers_lm_suppressors_avg_zscore_l$OrfRep), ]
  interaction_scores_deletion_enhancers_avg_zscore_suppressors_lm_l <-
    interaction_scores_deletion_enhancers_avg_zscore_suppressors_lm_l[
      !is.na(interaction_scores_deletion_enhancers_avg_zscore_suppressors_lm_l$OrfRep), ]
  interaction_scores_deletion_enhancers_lm_suppressors_avg_zscore_k <-
    interaction_scores_deletion_enhancers_lm_suppressors_avg_zscore_k[
      !is.na(interaction_scores_deletion_enhancers_lm_suppressors_avg_zscore_k$OrfRep), ]
  interaction_scores_deletion_enhancers_avg_zscore_suppressors_lm_k <-
    interaction_scores_deletion_enhancers_avg_zscore_suppressors_lm_k[
      !is.na(interaction_scores_deletion_enhancers_avg_zscore_suppressors_lm_k$OrfRep), ]
  write.csv(interaction_scores_deletion_enhancers_l,
    file.path(out_dir, "zscores_interaction_deletion_enhancers_l.csv"), row.names = FALSE)
  write.csv(interaction_scores_deletion_enhancers_k,
    file.path(out_dir, "zscores_interaction_deletion_enhancers_k.csv"), row.names = FALSE)
  write.csv(interaction_scores_deletion_suppressors_l,
    file.path(out_dir, "zscores_interaction_deletion_suppressors_l.csv"), row.names = FALSE)
  write.csv(interaction_scores_deletion_suppressors_k,
    file.path(out_dir, "zscores_interaction_deletion_suppressors_k.csv"), row.names = FALSE)
  write.csv(interaction_scores_deletion_enhancers_and_suppressors_l,
    file.path(out_dir, "zscores_interaction_deletion_enhancers_and_suppressors_l.csv"), row.names = FALSE)
  write.csv(interaction_scores_deletion_enhancers_and_suppressors_k,
    file.path(out_dir, "zscores_interaction_deletion_enhancers_and_suppressors_k.csv"), row.names = FALSE)
  write.csv(interaction_scores_deletion_enhancers_lm_suppressors_avg_zscore_l,
    file.path(out_dir, "zscores_interaction_suppressors_and_lm_enhancers_l.csv"), row.names = FALSE)
  write.csv(interaction_scores_deletion_enhancers_avg_zscore_suppressors_lm_l,
    file.path(out_dir, "zscores_interaction_enhancers_and_lm_suppressors_l.csv"), row.names = FALSE)
  write.csv(interaction_scores_deletion_enhancers_lm_suppressors_avg_zscore_k,
    file.path(out_dir, "zscores_interaction_suppressors_and_lm_enhancers_k.csv"), row.names = FALSE)
  write.csv(interaction_scores_deletion_enhancers_avg_zscore_suppressors_lm_k,
    file.path(out_dir, "zscores_interaction_enhancers_and_lm_suppressors_k.csv"), row.names = FALSE)
  
  # Get enhancers and suppressors for linear regression
  interaction_scores_deletion_enhancers_l_lm <-
    interaction_scores[interaction_scores$z_lm_l >= 2, ]
  interaction_scores_deletion_enhancers_k_lm <-
    interaction_scores[interaction_scores$z_lm_k <= -2, ]
  interaction_scores_deletion_suppressors_l_lm <-
    interaction_scores[interaction_scores$z_lm_l <= -2, ]
  interaction_scores_deletion_suppressors_k_lm <-
    interaction_scores[interaction_scores$z_lm_k >= 2, ]
  interaction_scores_deletion_enhancers_and_suppressors_l_lm <-
    interaction_scores[interaction_scores$z_lm_l >= 2 | interaction_scores$z_lm_l <= -2, ]
  interaction_scores_deletion_enhancers_and_suppressors_k_lm <-
    interaction_scores[interaction_scores$z_lm_k >= 2 | interaction_scores$z_lm_k <= -2, ]
  interaction_scores_deletion_enhancers_l_lm <-
    interaction_scores_deletion_enhancers_l_lm[
      !is.na(interaction_scores_deletion_enhancers_l_lm$OrfRep), ]
  interaction_scores_deletion_enhancers_k_lm <-
    interaction_scores_deletion_enhancers_k_lm[
      !is.na(interaction_scores_deletion_enhancers_k_lm$OrfRep), ]
  interaction_scores_deletion_suppressors_l_lm <-
    interaction_scores_deletion_suppressors_l_lm[
      !is.na(interaction_scores_deletion_suppressors_l_lm$OrfRep), ]
  interaction_scores_deletion_suppressors_k_lm <-
    interaction_scores_deletion_suppressors_k_lm[
      !is.na(interaction_scores_deletion_suppressors_k_lm$OrfRep), ]
  interaction_scores_deletion_enhancers_and_suppressors_l_lm <-
    interaction_scores_deletion_enhancers_and_suppressors_l_lm[
      !is.na(interaction_scores_deletion_enhancers_and_suppressors_l_lm$OrfRep), ]
  interaction_scores_deletion_enhancers_and_suppressors_k_lm <-
    interaction_scores_deletion_enhancers_and_suppressors_k_lm[
      !is.na(interaction_scores_deletion_enhancers_and_suppressors_k_lm$OrfRep), ]
  
  write.csv(interaction_scores_deletion_enhancers_l_lm,
    file.path(out_dir, "zscores_interaction_deletion_enhancers_l_lm.csv"), row.names = FALSE)
  write.csv(interaction_scores_deletion_enhancers_k_lm,
    file.path(out_dir, "zscores_interaction_deletion_enhancers_k_lm.csv"), row.names = FALSE)
  write.csv(interaction_scores_deletion_suppressors_l_lm,
    file.path(out_dir, "zscores_interaction_deletion_suppressors_l_lm.csv"), row.names = FALSE)
  write.csv(interaction_scores_deletion_suppressors_k_lm,
    file.path(out_dir, "zscores_interaction_deletion_suppressors_k_lm.csv"), row.names = FALSE)
  write.csv(interaction_scores_deletion_enhancers_and_suppressors_l_lm,
    file.path(out_dir, "zscores_interaction_deletion_enhancers_and_suppressors_l_lm.csv"), row.names = FALSE)
  write.csv(interaction_scores_deletion_enhancers_and_suppressors_k_lm,
    file.path(out_dir, "zscores_interaction_deletion_enhancers_and_suppressors_k_lm.csv"), row.names = FALSE)
  # write.csv(Labels, study_info_file, row.names = FALSE)
  # write.table(Labels, file.path("../Code/StudyInfo.txt"), sep = "\t", row.names = FALSE)
  
  for (i in 1:num_genes) {
    gene_sel <- unique(interaction_scores$OrfRep)[i]
    x_z_calculations <- x_stats_interaction_all[x_stats_interaction_all$OrfRep == gene_sel, ]
    df_int_scores <- interaction_scores[interaction_scores$OrfRep == gene_sel, ]
  
    p_l[[i]] <- ggplot(x_z_calculations, aes(conc_num_factor, delta_l)) +
      geom_point() + geom_smooth(method = "lm", formula = y ~ x, se = FALSE) +
      coord_cartesian(ylim = c(-65, 65)) +
      geom_errorbar(aes(ymin = 0 - (2 * WT_sd_l), ymax = 0 + (2 * WT_sd_l)), alpha = 0.3) +
      ggtitle(paste(x_z_calculations$OrfRep[1], x_z_calculations$Gene[1], sep = "      ")) +
      annotate("text", x = 1, y = 45, label = paste("ZShift =", round(df_int_scores$z_shift_l, 2))) +
      scale_color_discrete(guide = FALSE) +
      # annotate("text", x = 1, y = 35, label = paste("Avg ZScore =", round(df_int_scores$avg_zscore_l, 2))) +
      annotate("text", x = 1, y = 25, label = paste("Z lm Score =", round(df_int_scores$z_lm_l, 2))) +
      annotate("text", x = 1, y = -25, label = paste("NG =", df_int_scores$NG)) +
      annotate("text", x = 1, y = -35, label = paste("DB =", df_int_scores$DB)) +
      annotate("text", x = 1, y = -45, label = paste("SM =", df_int_scores$SM)) +
      scale_x_continuous(
        name = unique(df$Drug[1]),
        breaks = unique(x_z_calculations$conc_num_factor),
        labels = unique(as.character(x_z_calculations$conc_num))) +
      scale_y_continuous(breaks = c(-60, -50, -40, -30, -20, -10, 0, 10, 20, 30, 40, 50, 60)) +
      theme_publication()
    
    p_k[[i]] <- ggplot(x_z_calculations, aes(conc_num_factor, delta_k)) +
      geom_point() + geom_smooth(method = "lm", formula = y ~ x, se = FALSE) +
      coord_cartesian(ylim = c(-65, 65)) +
      geom_errorbar(aes(ymin = 0 - (2 * WT_sd_k), ymax = 0 + (2 * WT_sd_k)), alpha = 0.3) +
      ggtitle(paste(x_z_calculations$OrfRep[1], x_z_calculations$Gene[1], sep = "      ")) +
      annotate("text", x = 1, y = 45, label = paste("ZShift =", round(df_int_scores$z_shift_k, 2))) +
      scale_color_discrete(guide = FALSE) +
      # annotate("text", x = 1, y = 35, label = paste("Avg ZScore =", round(df_int_scores$avg_zscore_k, 2))) +
      annotate("text", x = 1, y = 25, label = paste("Z lm Score =", round(df_int_scores$z_lm_k, 2))) +
      annotate("text", x = 1, y = -25, label = paste("NG =", df_int_scores$NG)) +
      annotate("text", x = 1, y = -35, label = paste("DB =", df_int_scores$DB)) +
      annotate("text", x = 1, y = -45, label = paste("SM =", df_int_scores$SM)) +
      scale_x_continuous(
        name = unique(df$Drug[1]),
        breaks = unique(x_z_calculations$conc_num_factor),
        labels = unique(as.character(x_z_calculations$conc_num))) +
      scale_y_continuous(breaks = c(-60, -50, -40, -30, -20, -10, 0, 10, 20, 30, 40, 50, 60)) +
      theme_publication()
    
    p_r[[i]] <- ggplot(x_z_calculations, aes(conc_num_factor, delta_r)) +
      geom_point() + geom_smooth(method = "lm", formula = y ~ x, se = FALSE) +
      coord_cartesian(ylim = c(-0.65, 0.65)) +
      geom_errorbar(aes(ymin = 0 - (2 * WT_sd_r), ymax = 0 + (2 * WT_sd_r)), alpha = 0.3) +
      ggtitle(paste(x_z_calculations$OrfRep[1], x_z_calculations$Gene[1], sep = "      ")) +
      annotate("text", x = 1, y = 0.45, label = paste("ZShift =", round(df_int_scores$z_shift_r, 2))) +
      scale_color_discrete(guide = FALSE) +
      # annotate("text", x = 1, y = 0.35, label = paste("Avg ZScore =", round(df_int_scores$avg_zscore_r, 2))) +
      annotate("text", x = 1, y = 0.25, label = paste("Z lm Score =", round(df_int_scores$z_lm_r, 2))) +
      annotate("text", x = 1, y = -0.25, label = paste("NG =", df_int_scores$NG)) +
      annotate("text", x = 1, y = -0.35, label = paste("DB =", df_int_scores$DB)) +
      annotate("text", x = 1, y = -0.45, label = paste("SM =", df_int_scores$SM)) +
      scale_x_continuous(
        name = unique(df$Drug[1]),
        breaks = unique(x_z_calculations$conc_num_factor),
        labels = unique(as.character(x_z_calculations$conc_num))) +
      scale_y_continuous(breaks = c(-0.6, -0.4, -0.2, 0, 0.2, 0.4, 0.6)) +
      theme_publication()
    
    p_auc[[i]] <- ggplot(x_z_calculations, aes(conc_num_factor, delta_auc)) +
      geom_point() + geom_smooth(method = "lm", formula = y ~ x, se = FALSE) +
      coord_cartesian(ylim = c(-6500, 6500)) +
      geom_errorbar(aes(ymin = 0 - (2 * WT_sd_auc), ymax = 0 + (2 * WT_sd_auc)), alpha = 0.3) +
      ggtitle(paste(x_z_calculations$OrfRep[1], x_z_calculations$Gene[1], sep = "      ")) +
      annotate("text", x = 1, y = 4500, label = paste("ZShift =", round(df_int_scores$z_shift_auc, 2))) +
      scale_color_discrete(guide = FALSE) +
      # annotate("text", x = 1, y = 3500, label = paste("Avg ZScore =", round(df_int_scores$avg_zscore_auc, 2))) +
      annotate("text", x = 1, y = 2500, label = paste("Z lm Score =", round(df_int_scores$z_lm_auc, 2))) +
      annotate("text", x = 1, y = -2500, label = paste("NG =", df_int_scores$NG)) +
      annotate("text", x = 1, y = -3500, label = paste("DB =", df_int_scores$DB)) +
      annotate("text", x = 1, y = -4500, label = paste("SM =", df_int_scores$SM)) +
      scale_x_continuous(
        name = unique(df$Drug[1]),
        breaks = unique(x_z_calculations$conc_num_factor),
        labels = unique(as.character(x_z_calculations$conc_num))) +
      scale_y_continuous(breaks = c(-6000, -5000, -4000, -3000, -2000, -1000, 0, 1000, 2000, 3000, 4000, 5000, 6000)) +
      theme_publication()
    
    if (i == 1) {
      x_stats_interaction_all_final <- x_z_calculations
    }
    if (i > 1) {
      x_stats_interaction_all_final <- rbind(x_stats_interaction_all_final, x_z_calculations)
    }
  }
  print("Pass Int ggplot loop")
  write.csv(x_stats_interaction_all_final, file.path(out_dir, "zscore_calculations.csv"), row.names = FALSE)
  
  blank <- ggplot(df2_rf) + geom_blank()
  
  pdf(file.path(out_dir, "interaction_plots.pdf"), width = 16, height = 16, onefile = TRUE)
  
  x_stats_df2_rf <- ddply(
    df2_rf,
    c("conc_num", "conc_num_factor"),
    summarise,
    mean_l = mean(l, na.rm = TRUE),
    median_l = median(l, na.rm = TRUE),
    max_l = max(l, na.rm = TRUE),
    min_l = min(l, na.rm = TRUE),
    sd_l = sd(l, na.rm = TRUE),
    mean_k = mean(k, na.rm = TRUE),
    median_k = median(k, na.rm = TRUE),
    max_k = max(k, na.rm = TRUE),
    min_k = min(k, na.rm = TRUE),
    sd_k = sd(k, na.rm = TRUE),
    mean_r = mean(r, na.rm = TRUE),
    median_r = median(r, na.rm = TRUE),
    max_r = max(r, na.rm = TRUE),
    min_r = min(r, na.rm = TRUE),
    sd_r = sd(r, na.rm = TRUE),
    mean_auc = mean(auc, na.rm = TRUE),
    median_auc = median(auc, na.rm = TRUE),
    max_auc = max(auc, na.rm = TRUE),
    min_auc = min(auc, na.rm = TRUE),
    sd_auc = sd(auc, na.rm = TRUE),
    NG = sum(NG, na.rm = TRUE),
    DB = sum(DB, na.rm = TRUE),
    SM = sum(SM, na.rm = TRUE)
  )
  
  l_stats <- ggplot(df2_rf, aes(conc_num_factor, l)) + geom_point(position = "jitter", size = 1) +
    stat_summary(
      fun = mean,
      fun.min = function(x) mean(x) - sd(x),
      fun.max = function(x) mean(x) + sd(x),
      geom = "errorbar", color = "red") +
      stat_summary(fun = mean, geom = "point", color = "red") +
      scale_x_continuous(
        name = unique(df$Drug[1]),
        breaks = unique(df2_rf$conc_num_factor),
        labels = as.character(unique(df2_rf$conc_num))
      ) +
      ggtitle(paste(s, "Scatter RF for L with SD", sep = " ")) + coord_cartesian(ylim = c(0, 160)) +
      annotate("text", x = -0.25, y = 10, label = "NG") +
      annotate("text", x = -0.25, y = 5, label = "DB") +
      annotate("text", x = -0.25, y = 0, label = "SM") +
      annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = 10, label = x_stats_df2_rf$NG) +
      annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = 5, label = x_stats_df2_rf$DB) +
      annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = 0, label = x_stats_df2_rf$SM) +
      theme_publication()
  
  k_stats <- ggplot(df2_rf, aes(conc_num_factor, k)) + geom_point(position = "jitter", size = 1) +
    stat_summary(
      fun = mean,
      fun.min = function(x) mean(x) - sd(x),
      fun.max = function(x) mean(x) + sd(x),
      geom = "errorbar", color = "red") +
      stat_summary(fun = mean, geom = "point", color = "red") +
      scale_x_continuous(
        name = unique(df$Drug[1]),
        breaks = unique(df2_rf$conc_num_factor),
        labels = as.character(unique(df2_rf$conc_num))
      ) +
      ggtitle(paste(s, "Scatter RF for K with SD", sep = " ")) + coord_cartesian(ylim = c(-20, 160)) +
      annotate("text", x = -0.25, y = -5, label = "NG") +
      annotate("text", x = -0.25, y = -12.5, label = "DB") +
      annotate("text", x = -0.25, y = -20, label = "SM") +
      annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = -5, label = x_stats_df2_rf$NG) +
      annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = -12.5, label = x_stats_df2_rf$DB) +
      annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = -20, label = x_stats_df2_rf$SM) +
      theme_publication()
  
  r_stats <- ggplot(df2_rf, aes(conc_num_factor, r)) + geom_point(position = "jitter", size = 1) +
    stat_summary(
      fun = mean,
      fun.min = function(x) mean(x) - sd(x),
      fun.max = function(x) mean(x) + sd(x),
      geom = "errorbar", color = "red") +
      stat_summary(fun = mean, geom = "point", color = "red") +
      scale_x_continuous(
        name = unique(df$Drug[1]),
        breaks = unique(df2_rf$conc_num_factor),
        labels = as.character(unique(df2_rf$conc_num))
      ) +
      ggtitle(paste(s, "Scatter RF for r with SD", sep = " ")) + coord_cartesian(ylim = c(0, 1)) +
      annotate("text", x = -0.25, y = .9, label = "NG") +
      annotate("text", x = -0.25, y = .8, label = "DB") +
      annotate("text", x = -0.25, y = .7, label = "SM") +
      annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = .9, label = x_stats_df2_rf$NG) +
      annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = .8, label = x_stats_df2_rf$DB) +
      annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = .7, label = x_stats_df2_rf$SM) +
      theme_publication()
  
  auc_stats <- ggplot(df2_rf, aes(conc_num_factor, auc)) + geom_point(position = "jitter", size = 1) +
    stat_summary(
      fun = mean,
      fun.min = function(x) mean(x) - sd(x),
      fun.max = function(x) mean(x) + sd(x),
      geom = "errorbar", color = "red") +
      stat_summary(fun = mean, geom = "point", color = "red") +
      scale_x_continuous(
        name = unique(df$Drug[1]),
        breaks = unique(df2_rf$conc_num_factor),
        labels = as.character(unique(df2_rf$conc_num))
      ) +
      ggtitle(paste(s, "Scatter RF for auc with SD", sep = " ")) + coord_cartesian(ylim = c(0, 12500)) +
      annotate("text", x = -0.25, y = 11000, label = "NG") +
      annotate("text", x = -0.25, y = 10000, label = "DB") +
      annotate("text", x = -0.25, y = 9000, label = "SM") +
      annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = 11000, label = x_stats_df2_rf$NG) +
      annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = 10000, label = x_stats_df2_rf$DB) +
      annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = 9000, label = x_stats_df2_rf$SM) +
      theme_publication()
      
  l_stats_box <- ggplot(df2_rf, aes(as.factor(conc_num_factor), l)) +
    geom_boxplot() +
    scale_x_discrete(
      name = unique(df$Drug[1]),
      breaks = unique(df2_rf$conc_num_factor),
      labels = as.character(unique(df2_rf$conc_num))
    ) +
    ggtitle(paste(s, "Scatter RF for L with SD", sep = " ")) +
    coord_cartesian(ylim = c(0, 160)) +
    theme_publication()
  
  k_stats_box <- ggplot(df2_rf, aes(as.factor(conc_num_factor), k)) +
    geom_boxplot() +
    scale_x_discrete(
      name = unique(df$Drug[1]),
      breaks = unique(df2_rf$conc_num_factor),
      labels = as.character(unique(df2_rf$conc_num))
    ) +
    ggtitle(paste(s, "Scatter RF for K with SD", sep = " ")) +
    coord_cartesian(ylim = c(0, 130)) +
    theme_publication()
  
  r_stats_box <- ggplot(df2_rf, aes(as.factor(conc_num_factor), r)) +
    geom_boxplot() +
    scale_x_discrete(
      name = unique(df$Drug[1]),
      breaks = unique(df2_rf$conc_num_factor),
      labels = as.character(unique(df2_rf$conc_num))
    ) +
    ggtitle(paste(s, "Scatter RF for r with SD", sep = " ")) +
    coord_cartesian(ylim = c(0, 1)) +
    theme_publication()
  
  auc_stats_box <- ggplot(df2_rf, aes(as.factor(conc_num_factor), auc)) +
    geom_boxplot() +
    scale_x_discrete(
      name = unique(df$Drug[1]),
      breaks = unique(df2_rf$conc_num_factor),
      labels = as.character(unique(df2_rf$conc_num))
    ) +
    ggtitle(paste(s, "Scatter RF for auc with SD", sep = " ")) +
    coord_cartesian(ylim = c(0, 12500)) +
    theme_publication()
  
  grid.arrange(l_stats, k_stats, r_stats, auc_stats, ncol = 2, nrow = 2)
  grid.arrange(l_stats_box, k_stats_box, r_stats_box, auc_stats_box, ncol = 2, nrow = 2)
  
  # Plot the references
  # grid.arrange(p3, p3_k, p3_r, p4, p4_k, p4_r, p5, p5_k, p5_r, p6, p6_k, p6_r, ncol = 3, nrow = 4)
  # grid.arrange(p5, p5_k, p5_r, p6, p6_k, p6_r, ncol = 3, nrow = 2)
  
  # Loop for grid arrange 4x3
  j <- rep(1, ((num_genes) / 3) - 1)
  for (n in 1:length(j)) {
    j[n + 1] <- n * 3 + 1
  }

  # Loop for printing each plot
  num <- 0
  for (m in 1:(round((num_genes) / 3) - 1)) {
    num <- j[m]
    grid.arrange(p_l[[num]], p_k[[num]], p_r[[num]], p_auc[[num]],
    p_l[[num + 1]], p_k[[num + 1]], p_r[[num + 1]], p_auc[[num + 1]],
    p_l[[num + 2]], p_k[[num + 2]], p_r[[num + 2]], p_auc[[num + 2]], ncol = 4, nrow = 3)
    # grid.arrange(p_l[[364]], p_k[[364]], p_r[[364]],
    # p_l[[365]], p_k[[365]], p_r[[365]], p_l[[366]], p_k[[366]], p_r[[366]], ncol = 3, nrow = 3)
    # p1[[num + 3]], p_k[[num + 3]], p_r[[num + 3]], p1[[num + 4]], p_k[[num + 4]], p_r[[num + 4]]
  }

  if (num_genes != (num + 2)) {
    total_num <- num_genes - (num  +  2)
    if (total_num == 5) {
      grid.arrange(p_l[[num + 3]], p_k[[num + 3]], p_r[[num + 3]], p_auc[[num + 3]],
        p_l[[num + 4]], p_k[[num + 4]], p_r[[num + 4]], p_auc[[num + 4]],
        p_l[[num + 5]], p_k[[num + 5]], p_r[[num + 5]], p_auc[[num + 5]], ncol = 4, nrow = 3)

      grid.arrange(p_l[[num + 6]], p_k[[num + 6]], p_r[[num + 6]], p_auc[[num + 6]],
        p_l[[num + 7]], p_k[[num + 7]], p_r[[num + 7]], p_auc[[num + 7]],
        blank, blank, blank, blank, ncol = 4, nrow = 3)
    }
    if (total_num == 4) {
      grid.arrange(p_l[[num + 3]], p_k[[num + 3]], p_r[[num + 3]], p_auc[[num + 3]],
        p_l[[num + 4]], p_k[[num + 4]], p_r[[num + 4]], p_auc[[num + 4]],
        p_l[[num + 5]], p_k[[num + 5]], p_r[[num + 5]], p_auc[[num + 5]], ncol = 4, nrow = 3)

      grid.arrange(p_l[[num + 6]], p_k[[num + 6]], p_r[[num + 6]], p_auc[[num + 6]],
        blank, blank, blank, blank, blank, blank, blank, blank, ncol = 4, nrow = 3)
    }
    if (total_num == 3) {
      grid.arrange(p_l[[num + 3]], p_k[[num + 3]], p_r[[num + 3]], p_auc[[num + 3]],
        p_l[[num + 4]], p_k[[num + 4]], p_r[[num + 4]], p_auc[[num + 4]],
        p_l[[num + 5]], p_k[[num + 5]], p_r[[num + 5]], p_auc[[num + 5]], ncol = 4, nrow = 3)
      # grid.arrange(p_l[[num + 6]], p_k[[num + 6]], p_r[[num + 6]],
      # p_l[[num + 7]], p_k[[num + 7]], p_r[[num + 7]], blank, blank, blank, ncol = 3, nrow = 3)
    }
    if (total_num == 2) {
      grid.arrange(p_l[[num + 3]], p_k[[num + 3]], p_r[[num + 3]], p_auc[[num + 3]],
        p_l[[num + 4]], p_k[[num + 4]], p_r[[num + 4]], p_auc[[num + 4]],
        blank, blank, blank, blank, ncol = 4, nrow = 3)
      # grid.arrange(p_l[[num + 6]], p_k[[num + 6]], p_r[[num + 6]],
      # p_l[[num + 7]], p_k[[num + 7]], p_r[[num + 7]], blank, blank, blank, ncol = 3, nrow = 3)
    }
    if (total_num == 1) {
      grid.arrange(p_l[[num + 3]], p_k[[num + 3]], p_r[[num + 3]], p_auc[[num + 3]],
        blank, blank, blank, blank, blank, blank, blank, blank, ncol = 4, nrow = 3)
      # grid.arrange(p_l[[num + 6]], p_k[[num + 6]], p_r[[num + 6]],
      # p_l[[num + 7]], p_k[[num + 7]], p_r[[num + 7]], blank, blank, blank, ncol = 3, nrow = 3)
    }
  }

  dev.off()

  pdf(file.path(out_dir, "rf_interaction_plots.pdf"), width = 16, height = 16, onefile = TRUE)

  x_stats_df2_rf <- ddply(
    df2_rf,
    c("conc_num", "conc_num_factor"),
    summarise,
    mean_l = mean(l, na.rm = TRUE),
    median_l = median(l, na.rm = TRUE),
    max_l = max(l, na.rm = TRUE),
    min_l = min(l, na.rm = TRUE),
    sd_l = sd(l, na.rm = TRUE),
    mean_k = mean(k, na.rm = TRUE),
    median_k = median(k, na.rm = TRUE),
    max_k = max(k, na.rm = TRUE),
    min_k = min(k, na.rm = TRUE),
    sd_k = sd(k, na.rm = TRUE),
    mean_r = mean(r, na.rm = TRUE),
    median_r = median(r, na.rm = TRUE),
    max_r = max(r, na.rm = TRUE),
    min_r = min(r, na.rm = TRUE),
    sd_r = sd(r, na.rm = TRUE),
    mean_auc = mean(auc, na.rm = TRUE),
    median_auc = median(auc, na.rm = TRUE),
    max_auc = max(auc, na.rm = TRUE),
    min_auc = min(auc, na.rm = TRUE),
    sd_auc = sd(auc, na.rm = TRUE),
    NG = sum(NG, na.rm = TRUE),
    DB = sum(DB, na.rm = TRUE),
    SM = sum(SM, na.rm = TRUE)
  )
  
  l_stats <- ggplot(df2_rf, aes(conc_num_factor, l)) +
    geom_point(position = "jitter", size = 1) +
    stat_summary(
      fun = mean,
      fun.min = function(x) mean(x) - sd(x),
      fun.max = function(x) mean(x) + sd(x),
      geom = "errorbar", color = "red"
    ) +
    stat_summary(fun = mean, geom = "point", color = "red") +
    scale_x_continuous(
      name = unique(df$Drug[1]),
      breaks = unique(df2_rf$conc_num_factor),
      labels = as.character(unique(df2_rf$conc_num))
    ) +
    ggtitle(paste(s, "Scatter RF for L with SD", sep = " ")) +
    coord_cartesian(ylim = c(0, 130)) +
    annotate("text", x = -0.25, y = 10, label = "NG") +
    annotate("text", x = -0.25, y = 5, label = "DB") +
    annotate("text", x = -0.25, y = 0, label = "SM") +
    annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = 10, label = x_stats_df2_rf$NG) +
    annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = 5, label = x_stats_df2_rf$DB) +
    annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = 0, label = x_stats_df2_rf$SM) +
    theme_publication()
  
  k_stats <- ggplot(df2_rf, aes(conc_num_factor, k)) +
    geom_point(position = "jitter", size = 1) +
    stat_summary(
      fun = mean,
      fun.min = function(x) mean(x) - sd(x),
      fun.max = function(x) mean(x) + sd(x),
      geom = "errorbar", color = "red"
    ) +
    stat_summary(fun = mean, geom = "point", color = "red") +
    scale_x_continuous(
      name = unique(df$Drug[1]),
      breaks = unique(df2_rf$conc_num_factor),
      labels = as.character(unique(df2_rf$conc_num))
    ) +
    ggtitle(paste(s, "Scatter RF for K with SD", sep = " ")) +
    coord_cartesian(ylim = c(-20, 160)) +
    annotate("text", x = -0.25, y = -5, label = "NG") +
    annotate("text", x = -0.25, y = -12.5, label = "DB") +
    annotate("text", x = -0.25, y = -20, label = "SM") +
    annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = -5, label = x_stats_df2_rf$NG) +
    annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = -12.5, label = x_stats_df2_rf$DB) +
    annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = -20, label = x_stats_df2_rf$SM) +
    theme_publication()
  
  r_stats <- ggplot(df2_rf, aes(conc_num_factor, r)) +
    geom_point(position = "jitter", size = 1) +
    stat_summary(
      fun = mean,
      fun.min = function(x) mean(x) - sd(x),
      fun.max = function(x) mean(x) + sd(x),
      geom = "errorbar", color = "red"
    ) +
    stat_summary(fun = mean, geom = "point", color = "red") +
    scale_x_continuous(
      name = unique(df$Drug[1]),
      breaks = unique(df2_rf$conc_num_factor),
      labels = as.character(unique(df2_rf$conc_num))
    ) +
    ggtitle(paste(s, "Scatter RF for r with SD", sep = " ")) +
    coord_cartesian(ylim = c(0, 1)) +
    annotate("text", x = -0.25, y = .9, label = "NG") +
    annotate("text", x = -0.25, y = .8, label = "DB") +
    annotate("text", x = -0.25, y = .7, label = "SM") +
    annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = .9, label = x_stats_df2_rf$NG) +
    annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = .8, label = x_stats_df2_rf$DB) +
    annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = .7, label = x_stats_df2_rf$SM) +
    theme_publication()
  
  auc_stats <- ggplot(df2_rf, aes(conc_num_factor, auc)) +
    geom_point(position = "jitter", size = 1) +
    stat_summary(
      fun = mean,
      fun.min = function(x) mean(x) - sd(x),
      fun.max = function(x) mean(x) + sd(x),
      geom = "errorbar", color = "red"
    ) +
    stat_summary(fun = mean, geom = "point", color = "red") +
    scale_x_continuous(
      name = unique(df$Drug[1]),
      breaks = unique(df2_rf$conc_num_factor),
      labels = as.character(unique(df2_rf$conc_num))
    ) +
    ggtitle(paste(s, "Scatter RF for auc with SD", sep = " ")) +
    coord_cartesian(ylim = c(0, 12500)) +
    annotate("text", x = -0.25, y = 11000, label = "NG") +
    annotate("text", x = -0.25, y = 10000, label = "DB") +
    annotate("text", x = -0.25, y = 9000, label = "SM") +
    annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = 11000, label = x_stats_df2_rf$NG) +
    annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = 10000, label = x_stats_df2_rf$DB) +
    annotate("text", x = c(unique(df2_rf$conc_num_factor)), y = 9000, label = x_stats_df2_rf$SM) +
    theme_publication()
  
  l_stats_box <- ggplot(df2_rf, aes(as.factor(conc_num_factor), l)) +
    geom_boxplot() +
    scale_x_discrete(
      name = unique(df$Drug[1]),
      breaks = unique(df2_rf$conc_num_factor),
      labels = as.character(unique(df2_rf$conc_num))
    ) +
    ggtitle(paste(s, "Scatter RF for L with SD", sep = " ")) +
    coord_cartesian(ylim = c(0, 130)) +
    theme_publication()
  
  k_stats_box <- ggplot(df2_rf, aes(as.factor(conc_num_factor), k)) +
    geom_boxplot() +
    scale_x_discrete(
      name = unique(df$Drug[1]),
      breaks = unique(df2_rf$conc_num_factor),
      labels = as.character(unique(df2_rf$conc_num))
    ) +
    ggtitle(paste(s, "Scatter RF for K with SD", sep = " ")) +
    coord_cartesian(ylim = c(0, 160)) +
    theme_publication()
  
  r_stats_box <- ggplot(df2_rf, aes(as.factor(conc_num_factor), r)) +
    geom_boxplot() +
    scale_x_discrete(
      name = unique(df$Drug[1]),
      breaks = unique(df2_rf$conc_num_factor),
      labels = as.character(unique(df2_rf$conc_num))
    ) +
    ggtitle(paste(s, "Scatter RF for r with SD", sep = " ")) +
    coord_cartesian(ylim = c(0, 1)) +
    theme_publication()
  
  auc_stats_box <- ggplot(df2_rf, aes(as.factor(conc_num_factor), auc)) +
    geom_boxplot() +
    scale_x_discrete(
      name = unique(df$Drug[1]),
      breaks = unique(df2_rf$conc_num_factor),
      labels = as.character(unique(df2_rf$conc_num))
    ) +
    ggtitle(paste(s, "Scatter RF for auc with SD", sep = " ")) + coord_cartesian(ylim = c(12000, 0)) +
    theme_publication()
  
  grid.arrange(l_stats, k_stats, r_stats, auc_stats, ncol = 2, nrow = 2)
  grid.arrange(l_stats_box, k_stats_box, r_stats_box, auc_stats_box, ncol = 2, nrow = 2)
  
  # Plot the references
  # grid.arrange(p3, p3_k, p3_r, p4, p4_k, p4_r, p5, p5_k, p5_r, p6, p6_k, p6_r, ncol = 3, nrow = 4)
  # grid.arrange(p5, p5_k, p5_r, p6, p6_k, p6_r, ncol = 3, nrow = 2)
  
  # Loop for grid arrange 4x3
  j <- rep(1, ((num_genes_rf) / 3) - 1)
  for (n in 1:length(j)) {
    j[n + 1] <- n * 3 + 1
  }

  # Loop for printing each plot
  num <- 0
  for (m in 1:(round((num_genes_rf) / 3) - 1)) {
    num <- j[m]
    grid.arrange(p_rf_l[[num]], p_rf_k[[num]], p_rf_r[[num]], p_rf_auc[[num]],
      p_rf_l[[num + 1]], p_rf_k[[num + 1]], p_rf_r[[num + 1]], p_rf_auc[[num + 1]],
      p_rf_l[[num + 2]], p_rf_k[[num + 2]], p_rf_r[[num + 2]], p_rf_auc[[num + 2]], ncol = 4, nrow = 3)
    # grid.arrange(p_rf_l[[364]], p_rf_k[[364]], p_rf_r[[364]],
    # p_rf_l[[365]], p_rf_k[[365]], p_rf_r[[365]], p_rf_l[[366]], p_rf_k[[366]], p_rf_r[[366]], ncol = 3, nrow = 3)
    # p1[[num + 3]], p_rf_k[[num + 3]], p_rf_r[[num + 3]], p1[[num + 4]], p_rf_k[[num + 4]], p_rf_r[[num + 4]]
  }
  if (num_genes_rf != (num + 2)) {
    total_num <- num_genes_rf - (num + 2)

    if (total_num == 5) {
      grid.arrange(p_rf_l[[num + 3]], p_rf_k[[num + 3]], p_rf_r[[num + 3]], p_rf_auc[[num + 3]],
        p_rf_l[[num + 4]], p_rf_k[[num + 4]], p_rf_r[[num + 4]], p_rf_auc[[num + 4]],
        p_rf_l[[num + 5]], p_rf_k[[num + 5]], p_rf_r[[num + 5]], p_rf_auc[[num + 5]], ncol = 4, nrow = 3
      )
      grid.arrange(p_rf_l[[num + 6]], p_rf_k[[num + 6]], p_rf_r[[num + 6]], p_rf_auc[[num + 6]],
        p_rf_l[[num + 7]], p_rf_k[[num + 7]], p_rf_r[[num + 7]], p_rf_auc[[num + 7]],
        blank, blank, blank, blank, ncol = 4, nrow = 3
      )
    }

    if (total_num == 4) {
      grid.arrange(p_rf_l[[num + 3]], p_rf_k[[num + 3]], p_rf_r[[num + 3]], p_rf_auc[[num + 3]],
        p_rf_l[[num + 4]], p_rf_k[[num + 4]], p_rf_r[[num + 4]], p_rf_auc[[num + 4]],
        p_rf_l[[num + 5]], p_rf_k[[num + 5]], p_rf_r[[num + 5]], p_rf_auc[[num + 5]], ncol = 4, nrow = 3
      )
      grid.arrange(p_rf_l[[num + 6]], p_rf_k[[num + 6]], p_rf_r[[num + 6]], p_rf_auc[[num + 6]],
        blank, blank, blank, blank, blank, blank, blank, blank, ncol = 4, nrow = 3
      )
    }
    
    if (total_num == 3) {
      grid.arrange(p_rf_l[[num + 3]], p_rf_k[[num + 3]], p_rf_r[[num + 3]], p_rf_auc[[num + 3]],
        p_rf_l[[num + 4]], p_rf_k[[num + 4]], p_rf_r[[num + 4]], p_rf_auc[[num + 4]],
        p_rf_l[[num + 5]], p_rf_k[[num + 5]], p_rf_r[[num + 5]], p_rf_auc[[num + 5]], ncol = 4, nrow = 3
      )
      # grid.arrange(p_rf_l[[num + 6]], p_rf_k[[num + 6]], p_rf_r[[num + 6]],
      # p_rf_l[[num + 7]], p_rf_k[[num + 7]], p_rf_r[[num + 7]], blank, blank, blank, ncol = 3, nrow = 3)
    }
    
    if (total_num == 2) {
      grid.arrange(p_rf_l[[num + 3]], p_rf_k[[num + 3]], p_rf_r[[num + 3]], p_rf_auc[[num + 3]],
        p_rf_l[[num + 4]], p_rf_k[[num + 4]], p_rf_r[[num + 4]], p_rf_auc[[num + 4]],
        blank, blank, blank, blank, ncol = 4, nrow = 3
      )
      # grid.arrange(p_rf_l[[num + 6]], p_rf_k[[num + 6]], p_rf_r[[num + 6]],
      # p_rf_l[[num + 7]], p_rf_k[[num + 7]], p_rf_r[[num + 7]], blank, blank, blank, ncol = 3, nrow = 3)
    }
    
    if (total_num == 1) {
      grid.arrange(p_rf_l[[num + 3]], p_rf_k[[num + 3]], p_rf_r[[num + 3]], p_rf_auc[[num + 3]],
      blank, blank, blank, blank, blank, blank, blank, blank, ncol = 4, nrow = 3
      )
      # grid.arrange(p_rf_l[[num + 6]], p_rf_k[[num + 6]], p_rf_r[[num + 6]],
      # p_rf_l[[num + 7]], p_rf_k[[num + 7]], p_rf_r[[num + 7]], blank, blank, blank, ncol = 3, nrow = 3)
    }
  }

  dev.off()
  
  # Print rank plots for L and k gene interactions
  interaction_scores_adjust_missing <- interaction_scores
  interaction_scores_adjust_missing[is.na(interaction_scores_adjust_missing$avg_zscore_l), ]$avg_zscore_l <- 0.001
  interaction_scores_adjust_missing[is.na(interaction_scores_adjust_missing$avg_zscore_k), ]$avg_zscore_k <- 0.001
  interaction_scores_adjust_missing[is.na(interaction_scores_adjust_missing$avg_zscore_r), ]$avg_zscore_r <- 0.001
  interaction_scores_adjust_missing[is.na(interaction_scores_adjust_missing$avg_zscore_auc), ]$avg_zscore_auc <- 0.001
  
  interaction_scores_adjust_missing$l_rank <- NA
  interaction_scores_adjust_missing$k_rank <- NA
  interaction_scores_adjust_missing$r_rank <- NA
  interaction_scores_adjust_missing$auc_rank <- NA
  
  interaction_scores_adjust_missing$l_rank <- rank(interaction_scores_adjust_missing$avg_zscore_l)
  interaction_scores_adjust_missing$k_rank <- rank(interaction_scores_adjust_missing$avg_zscore_k)
  interaction_scores_adjust_missing$r_rank <- rank(interaction_scores_adjust_missing$avg_zscore_r)
  interaction_scores_adjust_missing$auc_rank <- rank(interaction_scores_adjust_missing$avg_zscore_auc)
  
  interaction_scores_adjust_missing[is.na(interaction_scores_adjust_missing$z_lm_l), ]$z_lm_l <- 0.001
  interaction_scores_adjust_missing[is.na(interaction_scores_adjust_missing$z_lm_k), ]$z_lm_k <- 0.001
  interaction_scores_adjust_missing[is.na(interaction_scores_adjust_missing$z_lm_r), ]$z_lm_r <- 0.001
  interaction_scores_adjust_missing[is.na(interaction_scores_adjust_missing$z_lm_auc), ]$z_lm_auc <- 0.001
  
  interaction_scores_adjust_missing$l_rank_lm <- NA
  interaction_scores_adjust_missing$k_rank_lm <- NA
  interaction_scores_adjust_missing$r_rank_lm <- NA
  interaction_scores_adjust_missing$auc_rank_lm <- NA
  
  interaction_scores_adjust_missing$l_rank_lm <- rank(interaction_scores_adjust_missing$z_lm_l)
  interaction_scores_adjust_missing$k_rank_lm <- rank(interaction_scores_adjust_missing$z_lm_k)
  interaction_scores_adjust_missing$r_rank_lm <- rank(interaction_scores_adjust_missing$z_lm_r)
  interaction_scores_adjust_missing$auc_rank_lm <- rank(interaction_scores_adjust_missing$z_lm_auc)
  
  # Rank plots
  rank_l_1sd <- ggplot(interaction_scores_adjust_missing, aes(l_rank, avg_zscore_l)) +
    ggtitle("Average Z score vs. Rank for L above 1SD") + xlab("Rank") + ylab("Avg Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (1), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-1), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-1, 1)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = 10, label = paste("Deletion Enhancers =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$avg_zscore_l >= 1, ])[1])) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = -10, label = paste("Deletion Suppressors =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$avg_zscore_l <= -1, ])[1])) +
    theme_publication()
  
  rank_l_2sd <- ggplot(interaction_scores_adjust_missing, aes(l_rank, avg_zscore_l)) +
    ggtitle("Average Z score vs. Rank for L above 2sd") + xlab("Rank") + ylab("Avg Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (2), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-2), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-2, 2)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = 10, label = paste("Deletion Enhancers =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$avg_zscore_l >= 2, ])[1])) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = -10, label = paste("Deletion Suppressors =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$avg_zscore_l <= -2, ])[1])) +
    theme_publication()
  
  rank_l_3sd <- ggplot(interaction_scores_adjust_missing, aes(l_rank, avg_zscore_l)) +
    ggtitle("Average Z score vs. Rank for L above 3sd") + xlab("Rank") + ylab("Avg Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (3), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-3), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-3, 3)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = 10, label = paste("Deletion Enhancers =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$avg_zscore_l >= 3, ])[1])) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = -10, label = paste("Deletion Suppressors =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$avg_zscore_l <= -3, ])[1])) +
    theme_publication()
  
  rank_k_1sd <- ggplot(interaction_scores_adjust_missing, aes(k_rank, avg_zscore_k)) +
    ggtitle("Average Z score vs. Rank for K above 1SD") + xlab("Rank") + ylab("Avg Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (1), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-1), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-1, 1)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = -10, label = paste("Deletion Enhancers =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$avg_zscore_k <= -1, ])[1])) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = 10, label = paste("Deletion Suppressors =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$avg_zscore_k >= 1, ])[1])) +
    theme_publication()
  
  rank_k_2sd <- ggplot(interaction_scores_adjust_missing, aes(k_rank, avg_zscore_k)) +
    ggtitle("Average Z score vs. Rank for K above 2sd") + xlab("Rank") + ylab("Avg Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (2), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-2), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-2, 2)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = -10, label = paste("Deletion Enhancers =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$avg_zscore_k <= -2, ])[1])) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = 10, label = paste("Deletion Suppressors =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$avg_zscore_k >= 2, ])[1])) +
    theme_publication()
  
  rank_k_3sd <- ggplot(interaction_scores_adjust_missing, aes(k_rank, avg_zscore_k)) +
    ggtitle("Average Z score vs. Rank for K above 3sd") + xlab("Rank") + ylab("Avg Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (3), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-3), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-3, 3)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = -10, label = paste("Deletion Enhancers =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$avg_zscore_k <= -3, ])[1])) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = 10, label = paste("Deletion Suppressors =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$avg_zscore_k >= 3, ])[1])) +
    theme_publication()
  
  rank_l_1sd_notext <- ggplot(interaction_scores_adjust_missing, aes(l_rank, avg_zscore_l)) +
    ggtitle("Average Z score vs. Rank for L above 1SD") + xlab("Rank") + ylab("Avg Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (1), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-1), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-1, 1)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  rank_l_2sd_notext <- ggplot(interaction_scores_adjust_missing, aes(l_rank, avg_zscore_l)) +
    ggtitle("Average Z score vs. Rank for L above 2sd") + xlab("Rank") + ylab("Avg Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (2), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-2), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-2, 2)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  rank_l_3sd_notext <- ggplot(interaction_scores_adjust_missing, aes(l_rank, avg_zscore_l)) +
    ggtitle("Average Z score vs. Rank for L above 3sd") + xlab("Rank") + ylab("Avg Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (3), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-3), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-3, 3)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  rank_k_1sd_notext <- ggplot(interaction_scores_adjust_missing, aes(k_rank, avg_zscore_k)) +
    ggtitle("Average Z score vs. Rank for K above 1SD") + xlab("Rank") + ylab("Avg Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (1), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-1), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-1, 1)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  rank_k_2sd_notext <- ggplot(interaction_scores_adjust_missing, aes(k_rank, avg_zscore_k)) +
    ggtitle("Average Z score vs. Rank for K above 2sd") + xlab("Rank") + ylab("Avg Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (2), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-2), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-2, 2)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  rank_k_3sd_notext <- ggplot(interaction_scores_adjust_missing, aes(k_rank, avg_zscore_k)) +
    ggtitle("Average Z score vs. Rank for K above 3sd") + xlab("Rank") + ylab("Avg Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (3), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-3), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-3, 3)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  pdf(file.path(out_dir, "rank_plots.pdf"), width = 18, height = 12, onefile = TRUE)
  
  grid.arrange(
    rank_l_1sd,
    rank_l_2sd,
    rank_l_3sd,
    rank_k_1sd,
    rank_k_2sd,
    rank_k_3sd,
    ncol = 3,
    nrow = 2
  )

  grid.arrange(
    rank_l_1sd_notext,
    rank_l_2sd_notext,
    rank_l_3sd_notext,
    rank_k_1sd_notext,
    rank_k_2sd_notext,
    rank_k_3sd_notext,
    ncol = 3,
    nrow = 2
  )
  
  dev.off()
  
  rank_l_1sd_lm <- ggplot(interaction_scores_adjust_missing, aes(l_rank_lm, z_lm_l)) +
    ggtitle("Interaction Z score vs. Rank for L above 1SD") + xlab("Rank") + ylab("Int Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (1), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-1), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-1, 1)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = 10, label = paste("Deletion Enhancers =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$z_lm_l >= 1, ])[1])) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = -10, label = paste("Deletion Suppressors =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$z_lm_l <= -1, ])[1])) +
    theme_publication()
  
  rank_l_2sd_lm <- ggplot(interaction_scores_adjust_missing, aes(l_rank_lm, z_lm_l)) +
    ggtitle("Interaction Z score vs. Rank for L above 2sd") + xlab("Rank") + ylab("Int Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (2), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-2), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-2, 2)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = 10, label = paste("Deletion Enhancers =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$z_lm_l >= 2, ])[1])) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = -10, label = paste("Deletion Suppressors =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$z_lm_l <= -2, ])[1])) +
    theme_publication()
  
  rank_l_3sd_lm <- ggplot(interaction_scores_adjust_missing, aes(l_rank_lm, z_lm_l)) +
    ggtitle("Interaction Z score vs. Rank for L above 3sd") + xlab("Rank") + ylab("Int Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (3), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-3), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-3, 3)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = 10, label = paste("Deletion Enhancers =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$z_lm_l >= 3, ])[1])) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = -10, label = paste("Deletion Suppressors =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$z_lm_l <= -3, ])[1])) +
    theme_publication()
  
  rank_k_1sd_lm <- ggplot(interaction_scores_adjust_missing, aes(k_rank_lm, z_lm_k)) +
    ggtitle("Interaction Z score vs. Rank for K above 1SD") + xlab("Rank") + ylab("Int Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (1), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-1), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-1, 1)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = -10, label = paste("Deletion Enhancers =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$z_lm_k <= -1, ])[1])) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = 10, label = paste("Deletion Suppressors =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$z_lm_k >= 1, ])[1])) +
    theme_publication()
  
  rank_k_2sd_lm <- ggplot(interaction_scores_adjust_missing, aes(k_rank_lm, z_lm_k)) +
    ggtitle("Interaction Z score vs. Rank for K above 2sd") + xlab("Rank") + ylab("Int Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (2), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-2), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-2, 2)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = -10, label = paste("Deletion Enhancers =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$z_lm_k <= -2, ])[1])) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = 10, label = paste("Deletion Suppressors =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$z_lm_k >= 2, ])[1])) +
    theme_publication()
  
  rank_k_3sd_lm <- ggplot(interaction_scores_adjust_missing, aes(k_rank_lm, z_lm_k)) +
    ggtitle("Interaction Z score vs. Rank for K above 3sd") + xlab("Rank") + ylab("Int Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (3), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-3), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-3, 3)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = -10, label = paste("Deletion Enhancers =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$z_lm_k <= -3, ])[1])) +
    annotate("text", x = (dim(interaction_scores_adjust_missing)[1] / 2), y = 10, label = paste("Deletion Suppressors =",
      dim(interaction_scores_adjust_missing[interaction_scores_adjust_missing$z_lm_k >= 3, ])[1])) +
    theme_publication()
  
  rank_l_1sd_notext_lm <- ggplot(interaction_scores_adjust_missing, aes(l_rank_lm, z_lm_l)) +
    ggtitle("Interaction Z score vs. Rank for L above 1SD") + xlab("Rank") + ylab("Int Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (1), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-1), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-1, 1)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  rank_l_2sd_notext_lm <- ggplot(interaction_scores_adjust_missing, aes(l_rank_lm, z_lm_l)) +
    ggtitle("Interaction Z score vs. Rank for L above 2sd") + xlab("Rank") + ylab("Int Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (2), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-2), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-2, 2)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  rank_l_3sd_notext_lm <- ggplot(interaction_scores_adjust_missing, aes(l_rank_lm, z_lm_l)) +
    ggtitle("Interaction Z score vs. Rank for L above 3sd") + xlab("Rank") + ylab("Int Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (3), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-3), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-3, 3)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  rank_k_1sd_notext_lm <- ggplot(interaction_scores_adjust_missing, aes(k_rank_lm, z_lm_k)) +
    ggtitle("Interaction Z score vs. Rank for K above 1SD") + xlab("Rank") + ylab("Int Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (1), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-1), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-1, 1)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  rank_k_2sd_notext_lm <- ggplot(interaction_scores_adjust_missing, aes(k_rank_lm, z_lm_k)) +
    ggtitle("Interaction Z score vs. Rank for K above 2sd") + xlab("Rank") + ylab("Int Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (2), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-2), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-2, 2)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  rank_k_3sd_notext_lm <- ggplot(interaction_scores_adjust_missing, aes(k_rank_lm, z_lm_k)) +
    ggtitle("Interaction Z score vs. Rank for K above 3sd") + xlab("Rank") + ylab("Int Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (3), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-3), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-3, 3)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  pdf(file.path(out_dir, "rank_plots_lm.pdf"), width = 18, height = 12, onefile = TRUE)
  
  grid.arrange(
    rank_l_1sd_lm,
    rank_l_2sd_lm,
    rank_l_3sd_lm,
    rank_k_1sd_lm,
    rank_k_2sd_lm,
    rank_k_3sd_lm,
    ncol = 3, nrow = 2
  )
  
  grid.arrange(
    rank_l_1sd_notext_lm,
    rank_l_2sd_notext_lm,
    rank_l_3sd_notext_lm,
    rank_k_1sd_notext_lm,
    rank_k_2sd_notext_lm,
    rank_k_3sd_notext_lm,
    ncol = 3, nrow = 2
  )
  
  dev.off()
  
  df_na_rm <- interaction_scores[!is.na(interaction_scores$z_lm_l) | !is.na(interaction_scores$avg_zscore_l), ]
  
  # Find overlaps
  df_na_rm$Overlap <- "No Effect"
  try(df_na_rm[df_na_rm$z_lm_l >= 2 & df_na_rm$avg_zscore_l >= 2, ]$Overlap <- "Deletion Enhancer Both")
  try(df_na_rm[df_na_rm$z_lm_l <= -2 & df_na_rm$avg_zscore_l <= -2, ]$Overlap <- "Deletion Suppressor Both")
  try(df_na_rm[df_na_rm$z_lm_l >= 2 & df_na_rm$avg_zscore_l <= 2, ]$Overlap <- "Deletion Enhancer lm only")
  try(df_na_rm[df_na_rm$z_lm_l <= 2 & df_na_rm$avg_zscore_l >= 2, ]$Overlap <- "Deletion Enhancer Avg Zscore only")
  try(df_na_rm[df_na_rm$z_lm_l <= -2 & df_na_rm$avg_zscore_l >= -2, ]$Overlap <- "Deletion Suppressor lm only")
  try(df_na_rm[df_na_rm$z_lm_l >= -2 & df_na_rm$avg_zscore_l <= -2, ]$Overlap <- "Deletion Suppressor Avg Zscore only")
  try(df_na_rm[df_na_rm$z_lm_l >= 2 & df_na_rm$avg_zscore_l <= -2, ]$Overlap <- "Deletion Enhancer lm, Deletion Suppressor Avg Z score")
  try(df_na_rm[df_na_rm$z_lm_l <= -2 & df_na_rm$avg_zscore_l >= 2, ]$Overlap <- "Deletion Suppressor lm, Deletion Enhancer Avg Z score")
  
  # Get the linear model info and the r-squared value for all CPPs in results 1 vs results 2
  get_lm_l <- lm(df_na_rm$z_lm_l ~ df_na_rm$avg_zscore_l)
  l_lm <- summary(get_lm_l)
  
  get_lm_k <- lm(df_na_rm$z_lm_k ~ df_na_rm$avg_zscore_k)
  k_lm <- summary(get_lm_k)
  
  get_lm_r <- lm(df_na_rm$z_lm_r ~ df_na_rm$avg_zscore_r)
  r_lm <- summary(get_lm_r)
  
  get_lm_auc <- lm(df_na_rm$z_lm_auc ~ df_na_rm$avg_zscore_auc)
  auc_lm <- summary(get_lm_auc)
  
  pdf(file.path(out_dir, "avg_zscore_vs_lm_na_rm.pdf"), width = 16, height = 12, onefile = TRUE)
  
  print(ggplot(df_na_rm, aes(avg_zscore_l, z_lm_l)) +
    geom_point(aes(color = Overlap), shape = 3) +
    geom_smooth(method = "lm", color = 1) +
    ggtitle("Avg Zscore vs lm L") +
    geom_rect(aes(xmin = -2, xmax = 2, ymin = -2, ymax = 2), color = "grey20", size = 0.25, alpha = 0.1, inherit.aes = FALSE, fill = NA) +
    annotate("text", x = 0, y = 0, label = paste("R-squared = ", round(l_lm$r.squared, 2))) + theme_publication_legend_right())
  
  print(ggplot(df_na_rm, aes(avg_zscore_k, z_lm_k)) +
    geom_point(aes(color = Overlap), shape = 3) +
    geom_smooth(method = "lm", color = 1) +
    ggtitle("Avg Zscore vs lm K") +
    geom_rect(aes(xmin = -2, xmax = 2, ymin = -2, ymax = 2), color = "grey20", size = 0.25, alpha = 0.1, inherit.aes = FALSE, fill = NA) +
    annotate("text", x = 0, y = 0, label = paste("R-squared = ", round(k_lm$r.squared, 2))) + theme_publication_legend_right())
  
  print(ggplot(df_na_rm, aes(avg_zscore_r, z_lm_r)) +
    geom_point(aes(color = Overlap), shape = 3) +
    geom_smooth(method = "lm", color = 1) +
    ggtitle("Avg Zscore vs lm r") +
    geom_rect(aes(xmin = -2, xmax = 2, ymin = -2, ymax = 2), color = "grey20", size = 0.25, alpha = 0.1, inherit.aes = FALSE, fill = NA) +
    annotate("text", x = 0, y = 0, label = paste("R-squared = ", round(r_lm$r.squared, 2))) + theme_publication_legend_right())
  
  print(ggplot(df_na_rm, aes(avg_zscore_auc, z_lm_auc)) +
    geom_point(aes(color = Overlap), shape = 3) +
    geom_smooth(method = "lm", color = 1) +
    ggtitle("Avg Zscore vs lm AUC") +
    geom_rect(aes(xmin = -2, xmax = 2, ymin = -2, ymax = 2), color = "grey20", size = 0.25, alpha = 0.1, inherit.aes = FALSE, fill = NA) +
    annotate("text", x = 0, y = 0, label = paste("R-squared = ", round(auc_lm$r.squared, 2))) + theme_publication_legend_right())
  
  dev.off()
  
  lm_v_zscore_l <- ggplot(df_na_rm, aes(avg_zscore_l, z_lm_l, ORF = OrfRep, Gene = Gene, NG = NG, SM = SM, DB = DB)) +
    geom_point(aes(color = Overlap), shape = 3) +
    geom_smooth(method = "lm", color = 1) +  ggtitle("Avg Zscore vs lm L") +
    geom_rect(aes(xmin = -2, xmax = 2, ymin = -2, ymax = 2), color = "grey20", size = 0.25, alpha = 0.1, inherit.aes = FALSE, fill = NA) +
    annotate("text", x = 0, y = 0, label = paste("R-squared = ", round(l_lm$r.squared, 2))) + theme_publication_legend_right()
  
  pgg <- ggplotly(lm_v_zscore_l)
  plotly_path <- file.path(out_dir, "avg_zscore_vs_lm_na_rm.html")
  saveWidget(pgg, file = plotly_path, selfcontained = TRUE)
  
  df_na_rm$l_rank <- rank(df_na_rm$avg_zscore_l)
  df_na_rm$k_rank <- rank(df_na_rm$avg_zscore_k)
  df_na_rm$r_rank <- rank(df_na_rm$avg_zscore_r)
  df_na_rm$auc_rank <- rank(df_na_rm$avg_zscore_auc)
  df_na_rm$l_rank_lm <- rank(df_na_rm$z_lm_l)
  df_na_rm$k_rank_lm <- rank(df_na_rm$z_lm_k)
  df_na_rm$r_rank_lm <- rank(df_na_rm$z_lm_r)
  df_na_rm$auc_rank_lm <- rank(df_na_rm$z_lm_auc)
  
  # Get the linear model info and the r-squared value for all CPPs in results 1 vs results 2
  get_lm_l2 <- lm(df_na_rm$l_rank_lm ~ df_na_rm$l_rank)
  l_lm2 <- summary(get_lm_l2)
  get_lm_k2 <- lm(df_na_rm$k_rank_lm ~ df_na_rm$k_rank)
  k_lm2 <- summary(get_lm_k2)
  get_lm_r2 <- lm(df_na_rm$r_rank_lm ~ df_na_rm$r_rank)
  r_lm2 <- summary(get_lm_r2)
  get_lm_auc2 <- lm(df_na_rm$auc_rank_lm ~ df_na_rm$auc_rank)
  auc_lm2 <- summary(get_lm_auc2)
  num_genes_na_rm2 <- (dim(df_na_rm)[1]) / 2
  
  pdf(file.path(out_dir, "avg_zscore_vs_lm_ranked_na_rm.pdf"), width = 16, height = 12, onefile = TRUE)
  
  print(
    ggplot(df_na_rm, aes(l_rank, l_rank_lm)) +
    geom_point(aes(color = Overlap), shape = 3) +
    geom_smooth(method = "lm", color = 1) +
    ggtitle("Rank Avg Zscore vs lm L") +
    annotate("text", x = num_genes_na_rm2, y = num_genes_na_rm2, label = paste("R-squared = ", round(l_lm2$r.squared, 2))) +
    theme_publication_legend_right()
  )
  
  print(
    ggplot(df_na_rm, aes(k_rank, k_rank_lm)) +
    geom_point(aes(color = Overlap), shape = 3) +
    geom_smooth(method = "lm", color = 1) +
    ggtitle("Rank Avg Zscore vs lm K") +
    annotate("text", x = num_genes_na_rm2, y = num_genes_na_rm2, label = paste("R-squared = ", round(k_lm2$r.squared, 2))) +
    theme_publication_legend_right()
  )
  
  print(
    ggplot(df_na_rm, aes(r_rank, r_rank_lm)) +
    geom_point(aes(color = Overlap), shape = 3) +
    geom_smooth(method = "lm", color = 1) +
    ggtitle("Rank Avg Zscore vs lm r") +
    annotate("text", x = num_genes_na_rm2, y = num_genes_na_rm2, label = paste("R-squared = ", round(r_lm2$r.squared, 2))) +
    theme_publication_legend_right()
  )
  
  print(
    ggplot(df_na_rm, aes(auc_rank, auc_rank_lm)) +
    geom_point(aes(color = Overlap), shape = 3) +
    geom_smooth(method = "lm", color = 1) +
    ggtitle("Rank of Avg Zscore vs lm AUC") +
    annotate("text", x = num_genes_na_rm2, y = num_genes_na_rm2, label = paste("R-squared = ", round(auc_lm2$r.squared, 2))) +
    theme_publication_legend_right()
  )
  
  dev.off()
  
  rank_l_1sd <- ggplot(df_na_rm, aes(l_rank, avg_zscore_l)) +
    ggtitle("Average Z score vs. Rank for L above 1SD") + xlab("Rank") + ylab("Avg Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (1), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-1), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-1, 1)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = 10, label = paste("Deletion Enhancers =", dim(df_na_rm[df_na_rm$avg_zscore_l >= 1, ])[1])) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = -10, label = paste("Deletion Suppressors =", dim(df_na_rm[df_na_rm$avg_zscore_l <= -1, ])[1])) +
    theme_publication()
  
  rank_l_2sd <- ggplot(df_na_rm, aes(l_rank, avg_zscore_l)) +
    ggtitle("Average Z score vs. Rank for L above 2sd") + xlab("Rank") + ylab("Avg Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (2), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-2), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-2, 2)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = 10, label = paste("Deletion Enhancers =", dim(df_na_rm[df_na_rm$avg_zscore_l >= 2, ])[1])) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = -10, label = paste("Deletion Suppressors =", dim(df_na_rm[df_na_rm$avg_zscore_l <= -2, ])[1])) +
    theme_publication()
  
  rank_l_3sd <- ggplot(df_na_rm, aes(l_rank, avg_zscore_l)) +
    ggtitle("Average Z score vs. Rank for L above 3sd") + xlab("Rank") + ylab("Avg Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (3), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-3), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-3, 3)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = 10, label = paste("Deletion Enhancers =", dim(df_na_rm[df_na_rm$avg_zscore_l >= 3, ])[1])) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = -10, label = paste("Deletion Suppressors =", dim(df_na_rm[df_na_rm$avg_zscore_l <= -3, ])[1])) +
    theme_publication()
  
  rank_k_1sd <- ggplot(df_na_rm, aes(k_rank, avg_zscore_k)) +
    ggtitle("Average Z score vs. Rank for K above 1SD") + xlab("Rank") + ylab("Avg Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (1), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-1), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-1, 1)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = -10, label = paste("Deletion Enhancers =", dim(df_na_rm[df_na_rm$avg_zscore_k <= -1, ])[1])) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = 10, label = paste("Deletion Suppressors =", dim(df_na_rm[df_na_rm$avg_zscore_k >= 1, ])[1])) +
    theme_publication()
  
  rank_k_2sd <- ggplot(df_na_rm, aes(k_rank, avg_zscore_k)) +
    ggtitle("Average Z score vs. Rank for K above 2sd") + xlab("Rank") + ylab("Avg Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (2), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-2), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-2, 2)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = -10, label = paste("Deletion Enhancers =", dim(df_na_rm[df_na_rm$avg_zscore_k <= -2, ])[1])) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = 10, label = paste("Deletion Suppressors =", dim(df_na_rm[df_na_rm$avg_zscore_k >= 2, ])[1])) +
    theme_publication()
  
  rank_k_3sd <- ggplot(df_na_rm, aes(k_rank, avg_zscore_k)) +
    ggtitle("Average Z score vs. Rank for K above 3sd") + xlab("Rank") + ylab("Avg Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (3), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-3), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-3, 3)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = -10, label = paste("Deletion Enhancers =", dim(df_na_rm[df_na_rm$avg_zscore_k <= -3, ])[1])) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = 10, label = paste("Deletion Suppressors =", dim(df_na_rm[df_na_rm$avg_zscore_k >= 3, ])[1])) +
    theme_publication()
  
  rank_l_1sd_notext <- ggplot(df_na_rm, aes(l_rank, avg_zscore_l)) +
    ggtitle("Average Z score vs. Rank for L above 1SD") + xlab("Rank") + ylab("Avg Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (1), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-1), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-1, 1)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  rank_l_2sd_notext <- ggplot(df_na_rm, aes(l_rank, avg_zscore_l)) +
    ggtitle("Average Z score vs. Rank for L above 2sd") + xlab("Rank") + ylab("Avg Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (2), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-2), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-2, 2)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  rank_l_3sd_notext <- ggplot(df_na_rm, aes(l_rank, avg_zscore_l)) +
    ggtitle("Average Z score vs. Rank for L above 3sd") + xlab("Rank") + ylab("Avg Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (3), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-3), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-3, 3)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  rank_k_1sd_notext <- ggplot(df_na_rm, aes(k_rank, avg_zscore_k)) +
    ggtitle("Average Z score vs. Rank for K above 1SD") + xlab("Rank") + ylab("Avg Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (1), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-1), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-1, 1)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  rank_k_2sd_notext <- ggplot(df_na_rm, aes(k_rank, avg_zscore_k)) +
    ggtitle("Average Z score vs. Rank for K above 2sd") + xlab("Rank") + ylab("Avg Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (2), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-2), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-2, 2)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  rank_k_3sd_notext <- ggplot(df_na_rm, aes(k_rank, avg_zscore_k)) +
    ggtitle("Average Z score vs. Rank for K above 3sd") + xlab("Rank") + ylab("Avg Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (3), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-3), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-3, 3)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  pdf(file.path(out_dir, "rank_plots_narm.pdf"), width = 18, height = 12, onefile = TRUE)
  
  grid.arrange(
    rank_l_1sd,
    rank_l_2sd,
    rank_l_3sd,
    rank_k_1sd,
    rank_k_2sd,
    rank_k_3sd,
    ncol = 3, nrow = 2
  )

  grid.arrange(
    rank_l_1sd_notext,
    rank_l_2sd_notext,
    rank_l_3sd_notext,
    rank_k_1sd_notext,
    rank_k_2sd_notext,
    rank_k_3sd_notext,
    ncol = 3, nrow = 2
  )
  
  dev.off()
  
  rank_l_1sd_lm <- ggplot(df_na_rm, aes(l_rank_lm, z_lm_l)) +
    ggtitle("Interaction Z score vs. Rank for L above 1SD") + xlab("Rank") + ylab("Int Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (1), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-1), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-1, 1)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = 10, label = paste("Deletion Enhancers =", dim(df_na_rm[df_na_rm$z_lm_l >= 1, ])[1])) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = -10, label = paste("Deletion Suppressors =", dim(df_na_rm[df_na_rm$z_lm_l <= -1, ])[1])) +
    theme_publication()
  
  rank_l_2sd_lm <- ggplot(df_na_rm, aes(l_rank_lm, z_lm_l)) +
    ggtitle("Interaction Z score vs. Rank for L above 2sd") + xlab("Rank") + ylab("Int Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (2), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-2), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-2, 2)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = 10, label = paste("Deletion Enhancers =", dim(df_na_rm[df_na_rm$z_lm_l >= 2, ])[1])) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = -10, label = paste("Deletion Suppressors =", dim(df_na_rm[df_na_rm$z_lm_l <= -2, ])[1])) +
    theme_publication()
  
  rank_l_3sd_lm <- ggplot(df_na_rm, aes(l_rank_lm, z_lm_l)) +
    ggtitle("Interaction Z score vs. Rank for L above 3sd") + xlab("Rank") + ylab("Int Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (3), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-3), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-3, 3)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = 10, label = paste("Deletion Enhancers =", dim(df_na_rm[df_na_rm$z_lm_l >= 3, ])[1])) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = -10, label = paste("Deletion Suppressors =", dim(df_na_rm[df_na_rm$z_lm_l <= -3, ])[1])) +
    theme_publication()
  
  rank_k_1sd_lm <- ggplot(df_na_rm, aes(k_rank_lm, z_lm_k)) +
    ggtitle("Interaction Z score vs. Rank for K above 1SD") + xlab("Rank") + ylab("Int Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (1), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-1), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-1, 1)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = -10, label = paste("Deletion Enhancers =", dim(df_na_rm[df_na_rm$z_lm_k <= -1, ])[1])) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = 10, label = paste("Deletion Suppressors =", dim(df_na_rm[df_na_rm$z_lm_k >= 1, ])[1])) +
    theme_publication()
  
  rank_k_2sd_lm <- ggplot(df_na_rm, aes(k_rank_lm, z_lm_k)) +
    ggtitle("Interaction Z score vs. Rank for K above 2sd") + xlab("Rank") + ylab("Int Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (2), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-2), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-2, 2)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = -10, label = paste("Deletion Enhancers =", dim(df_na_rm[df_na_rm$z_lm_k <= -2, ])[1])) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = 10, label = paste("Deletion Suppressors =", dim(df_na_rm[df_na_rm$z_lm_k >= 2, ])[1])) +
    theme_publication()
  
  rank_k_3sd_lm <- ggplot(df_na_rm, aes(k_rank_lm, z_lm_k)) +
    ggtitle("Interaction Z score vs. Rank for K above 3sd") + xlab("Rank") + ylab("Int Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (3), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-3), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-3, 3)) + geom_point(size = 0.1, shape = 3) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = -10, label = paste("Deletion Enhancers =", dim(df_na_rm[df_na_rm$z_lm_k <= -3, ])[1])) +
    annotate("text", x = (dim(df_na_rm)[1] / 2), y = 10, label = paste("Deletion Suppressors =", dim(df_na_rm[df_na_rm$z_lm_k >= 3, ])[1])) +
    theme_publication()
  
  rank_l_1sd_notext_lm <- ggplot(df_na_rm, aes(l_rank_lm, z_lm_l)) +
    ggtitle("Interaction Z score vs. Rank for L above 1SD") + xlab("Rank") + ylab("Int Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (1), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-1), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-1, 1)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  rank_l_2sd_notext_lm <- ggplot(df_na_rm, aes(l_rank_lm, z_lm_l)) +
    ggtitle("Interaction Z score vs. Rank for L above 2sd") + xlab("Rank") + ylab("Int Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (2), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-2), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-2, 2)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  rank_l_3sd_notext_lm <- ggplot(df_na_rm, aes(l_rank_lm, z_lm_l)) +
    ggtitle("Interaction Z score vs. Rank for L above 3sd") + xlab("Rank") + ylab("Int Z score L") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (3), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-3), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-3, 3)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  rank_k_1sd_notext_lm <- ggplot(df_na_rm, aes(k_rank_lm, z_lm_k)) +
    ggtitle("Interaction Z score vs. Rank for K above 1SD") + xlab("Rank") + ylab("Int Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (1), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-1), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-1, 1)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  rank_k_2sd_notext_lm <- ggplot(df_na_rm, aes(k_rank_lm, z_lm_k)) +
    ggtitle("Interaction Z score vs. Rank for K above 2sd") + xlab("Rank") + ylab("Int Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (2), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-2), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-2, 2)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  rank_k_3sd_notext_lm <- ggplot(df_na_rm, aes(k_rank_lm, z_lm_k)) +
    ggtitle("Interaction Z score vs. Rank for K above 3sd") + xlab("Rank") + ylab("Int Z score K") +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (3), ymax = Inf, fill = "#542788", alpha = 0.3) +
    annotate("rect", xmin = -Inf, xmax = Inf, ymin = (-3), ymax = -Inf, fill = "orange", alpha = 0.3) +
    geom_hline(yintercept = c(-3, 3)) + geom_point(size = 0.1, shape = 3) +
    theme_publication()
  
  pdf(file.path(out_dir, "rank_plots_lm_narm.pdf"), width = 18, height = 12, onefile = TRUE)
  
  grid.arrange(
    rank_l_1sd_lm,
    rank_l_2sd_lm,
    rank_l_3sd_lm,
    rank_k_1sd_lm,
    rank_k_2sd_lm,
    rank_k_3sd_lm,
    ncol = 3, nrow = 2
  )
  
  grid.arrange(
    rank_l_1sd_notext_lm,
    rank_l_2sd_notext_lm,
    rank_l_3sd_notext_lm,
    rank_k_1sd_notext_lm,
    rank_k_2sd_notext_lm,
    rank_k_3sd_notext_lm,
    ncol = 3, nrow = 2
  )
  
  dev.off()
}

# Get the linear model info and the r-squared value for all CPPs in results 1 vs results 2
get_lm_1 <- lm(df_na_rm$z_lm_k ~ df_na_rm$z_lm_l)
l_lm_1 <- summary(get_lm_1)
get_lm_2 <- lm(df_na_rm$z_lm_r ~ df_na_rm$z_lm_l)
l_lm_2 <- summary(get_lm_2)
get_lm_3 <- lm(df_na_rm$z_lm_auc ~ df_na_rm$z_lm_l)
l_lm_3 <- summary(get_lm_3)
get_lm_4 <- lm(df_na_rm$z_lm_r ~ df_na_rm$z_lm_k)
l_lm_4 <- summary(get_lm_4)
get_lm_5 <- lm(df_na_rm$z_lm_auc ~ df_na_rm$z_lm_k)
l_lm_5 <- summary(get_lm_5)
get_lm_6 <- lm(df_na_rm$z_lm_auc ~ df_na_rm$z_lm_r)
l_lm_6 <- summary(get_lm_6)

pdf(file.path(out_dir, "correlation_cpps.pdf"), width = 10, height = 7, onefile = TRUE)

ggplot(df_na_rm, aes(z_lm_l, z_lm_k)) +
  geom_point(shape = 3, color = "gray70") +
  geom_smooth(method = "lm", color = "tomato3") +
  ggtitle("Interaction L vs. Interaction K") +
  xlab("z-score L") +
  ylab("z-score K") +
  annotate("text", x = 0, y = 0, label = paste("R-squared = ", round(l_lm_1$r.squared, 3))) +
  theme_publication_legend_right() +
  theme(
    panel.grid.major = element_blank(),
    panel.grid.minor = element_blank(),
    axis.text.x = element_text(size = 16),
    axis.title.x = element_text(size = 18),
    axis.text.y = element_text(size = 16),
    axis.title.y = element_text(size = 18)
  )

ggplot(df_na_rm, aes(z_lm_l, z_lm_r)) +
  geom_point(shape = 3, color = "gray70") +
  geom_smooth(method = "lm", color = "tomato3") +
  ggtitle("Interaction L vs. Interaction r") +
  xlab("z-score L") +
  ylab("z-score r") +
  annotate("text", x = 0, y = 0, label = paste("R-squared = ", round(l_lm_2$r.squared, 3))) +
  theme_publication_legend_right() +
  theme(
    panel.grid.major = element_blank(),
    panel.grid.minor = element_blank(),
    axis.text.x = element_text(size = 16),
    axis.title.x = element_text(size = 18),
    axis.text.y = element_text(size = 16),
    axis.title.y = element_text(size = 18)
  )

ggplot(df_na_rm, aes(z_lm_l, z_lm_auc)) +
geom_point(shape = 3, color = "gray70") +
geom_smooth(method = "lm", color = "tomato3") +
ggtitle("Interaction L vs. Interaction AUC") +
xlab("z-score L") +
ylab("z-score AUC") +
annotate("text", x = 0, y = 0, label = paste("R-squared = ", round(l_lm_3$r.squared, 3))) +
theme_publication_legend_right() +
theme(
  panel.grid.major = element_blank(),
  panel.grid.minor = element_blank(),
  axis.text.x = element_text(size = 16),
  axis.title.x = element_text(size = 18),
  axis.text.y = element_text(size = 16),
  axis.title.y = element_text(size = 18)
)

ggplot(df_na_rm, aes(z_lm_k, z_lm_r)) +
  geom_point(shape = 3, color = "gray70") +
  geom_smooth(method = "lm", color = "tomato3") +
  ggtitle("Interaction K vs. Interaction r") +
  xlab("z-score K") +
  ylab("z-score r") +
  annotate("text", x = 0, y = 0, label = paste("R-squared = ", round(l_lm_4$r.squared, 3))) +
  theme_publication_legend_right() +
  theme(
    panel.grid.major = element_blank(),
    panel.grid.minor = element_blank(),
    axis.text.x = element_text(size = 16),
    axis.title.x = element_text(size = 18),
    axis.text.y = element_text(size = 16),
    axis.title.y = element_text(size = 18)
  )

ggplot(df_na_rm, aes(z_lm_k, z_lm_auc)) +
  geom_point(shape = 3, color = "gray70") +
  geom_smooth(method = "lm", color = "tomato3") +
  ggtitle("Interaction K vs. Interaction AUC") +
  xlab("z-score K") +
  ylab("z-score AUC") +
  annotate("text", x = 0, y = 0, label = paste("R-squared = ", round(l_lm_5$r.squared, 3))) +
  theme_publication_legend_right() +
  theme(
    panel.grid.major = element_blank(),
    panel.grid.minor = element_blank(),
    axis.text.x = element_text(size = 16),
    axis.title.x = element_text(size = 18),
    axis.text.y = element_text(size = 16),
    axis.title.y = element_text(size = 18)
  )

ggplot(df_na_rm, aes(z_lm_r, z_lm_auc)) +
  geom_point(shape = 3, color = "gray70") +
  geom_smooth(method = "lm", color = "tomato3") +
  ggtitle("Interaction r vs. Interaction AUC") +
  xlab("z-score r") +
  ylab("z-score AUC") +
  annotate("text", x = 0, y = 0, label = paste("R-squared = ", round(l_lm_6$r.squared, 3))) +
  theme_publication_legend_right() +
  theme(
    panel.grid.major = element_blank(),
    panel.grid.minor = element_blank(),
    axis.text.x = element_text(size = 16),
    axis.title.x = element_text(size = 18),
    axis.text.y = element_text(size = 16),
    axis.title.y = element_text(size = 18))

interaction_scores_rf2 <- interaction_scores_rf[!is.na(interaction_scores_rf$z_lm_l), ]

ggplot(df_na_rm, aes(z_lm_l, z_lm_k)) +
  geom_point(shape = 3, color = "gray70") +
  geom_point(data = interaction_scores_rf2, aes(z_lm_l, z_lm_k), color = "cyan") +
  ggtitle("Interaction L vs. Interaction K") +
  xlab("z-score L") +
  ylab("z-score K") +
  annotate("text", x = 0, y = 0, label = paste("R-squared = ", round(l_lm_1$r.squared, 3))) +
  theme_publication_legend_right() +
  theme(
    panel.grid.major = element_blank(),
    panel.grid.minor = element_blank(),
    axis.text.x = element_text(size = 16),
    axis.title.x = element_text(size = 18),
    axis.text.y = element_text(size = 16),
    axis.title.y = element_text(size = 18)
  )

ggplot(df_na_rm, aes(z_lm_l, z_lm_r)) +
  geom_point(shape = 3, color = "gray70") +
  geom_point(data = interaction_scores_rf2, aes(z_lm_l, z_lm_r), color = "cyan") +
  ggtitle("Interaction L vs. Interaction r") +
  xlab("z-score L") +
  ylab("z-score r") +
  annotate("text", x = 0, y = 0, label = paste("R-squared = ", round(l_lm_2$r.squared, 3))) +
  theme_publication_legend_right() +
  theme(
    panel.grid.major = element_blank(),
    panel.grid.minor = element_blank(),
    axis.text.x = element_text(size = 16),
    axis.title.x = element_text(size = 18),
    axis.text.y = element_text(size = 16),
    axis.title.y = element_text(size = 18)
  )

ggplot(df_na_rm, aes(z_lm_l, z_lm_auc)) +
  geom_point(shape = 3, color = "gray70") +
  geom_point(data = interaction_scores_rf2, aes(z_lm_l, z_lm_auc), color = "cyan") +
  ggtitle("Interaction L vs. Interaction AUC") +
  xlab("z-score L") +
  ylab("z-score AUC") +
  annotate("text", x = 0, y = 0, label = paste("R-squared = ", round(l_lm_3$r.squared, 3))) +
  theme_publication_legend_right() +
  theme(
    panel.grid.major = element_blank(),
    panel.grid.minor = element_blank(),
    axis.text.x = element_text(size = 16),
    axis.title.x = element_text(size = 18),
    axis.text.y = element_text(size = 16),
    axis.title.y = element_text(size = 18)
  )

ggplot(df_na_rm, aes(z_lm_k, z_lm_r)) +
  geom_point(shape = 3, color = "gray70") +
  geom_point(data = interaction_scores_rf2, aes(z_lm_k, z_lm_r), color = "cyan") +
  ggtitle("Interaction K vs. Interaction r") +
  xlab("z-score K") +
  ylab("z-score r") +
  annotate("text", x = 0, y = 0, label = paste("R-squared = ", round(l_lm_4$r.squared, 3))) +
  theme_publication_legend_right() +
  theme(
    panel.grid.major = element_blank(),
    panel.grid.minor = element_blank(),
    axis.text.x = element_text(size = 16),
    axis.title.x = element_text(size = 18),
    axis.text.y = element_text(size = 16),
    axis.title.y = element_text(size = 18)
  )

ggplot(df_na_rm, aes(z_lm_k, z_lm_auc)) +
  geom_point(shape = 3, color = "gray70") +
  geom_point(data = interaction_scores_rf2, aes(z_lm_k, z_lm_auc), color = "cyan") +
  ggtitle("Interaction K vs. Interaction AUC") +
  xlab("z-score K") +
  ylab("z-score AUC") +
  annotate("text", x = 0, y = 0, label = paste("R-squared = ", round(l_lm_5$r.squared, 3))) +
  theme_publication_legend_right() +
  theme(
    panel.grid.major = element_blank(),
    panel.grid.minor = element_blank(),
    axis.text.x = element_text(size = 16),
    axis.title.x = element_text(size = 18),
    axis.text.y = element_text(size = 16),
    axis.title.y = element_text(size = 18)
  )

ggplot(df_na_rm, aes(z_lm_r, z_lm_auc)) +
  geom_point(shape = 3, color = "gray70") +
  geom_point(data = interaction_scores_rf2, aes(z_lm_r, z_lm_auc), color = "cyan") +
  ggtitle("Interaction r vs. Interaction AUC") +
  xlab("z-score r") +
  ylab("z-score AUC") +
  annotate("text", x = 0, y = 0, label = paste("R-squared = ", round(l_lm_6$r.squared, 3))) +
  theme_publication_legend_right() +
  theme(
    panel.grid.major = element_blank(),
    panel.grid.minor = element_blank(),
    axis.text.x = element_text(size = 16),
    axis.title.x = element_text(size = 18),
    axis.text.y = element_text(size = 16),
    axis.title.y = element_text(size = 18)
  )

dev.off()

# write.csv(Labels, file.path("../Code/Parameters.csv"), row.names = FALSE)
timestamp()