Use precomputed se values with Bessel's correction

qhtcp-workflow/apps/r/calculate_interaction_zscores.R

@@ -189,7 +189,7 @@ calculate_summary_stats <- function(df, variables, group_vars) {
           max = ~ifelse(all(is.na(.)), NA, max(., na.rm = TRUE)),
           min = ~ifelse(all(is.na(.)), NA, min(., na.rm = TRUE)),
           sd = ~sd(., na.rm = TRUE),
-          se = ~sd(., na.rm = TRUE) / sqrt(N - 1) # TODO non-standard SE, needs explanation
+          se = ~sd(., na.rm = TRUE) / sqrt(N - 1) # Bessel's correction
         ),
         .names = "{.fn}_{.col}"
       ),
@@ -581,13 +581,13 @@ generate_scatter_plot <- function(plot, config) {
   # Add Error Bars if specified
   if (!is.null(config$error_bar) && config$error_bar && !is.null(config$y_var)) {
     y_mean_col <- paste0("mean_", config$y_var)
-    y_sd_col <- paste0("sd_", config$y_var)
+    y_se_col <- paste0("se_", config$y_var)
     
     plot <- plot +
       geom_errorbar(
         aes(
-          ymin = !!sym(y_mean_col) - !!sym(y_sd_col),
-          ymax = !!sym(y_mean_col) + !!sym(y_sd_col)
+          ymin = !!sym(y_mean_col) - !!sym(y_se_col),
+          ymax = !!sym(y_mean_col) + !!sym(y_se_col)
         ),
         alpha = 0.3
       )
@@ -676,7 +676,8 @@ generate_plate_analysis_plot_configs <- function(variables, stages = c("before",
         title = paste("Plate analysis by Drug Conc for", var, stage, "quality control"),
         error_bar = error_bar,
         color_var = "conc_num",
-        position = position
+        position = position,
+        size = 0.2
       )
       plots <- append(plots, list(config))
     }
@@ -760,7 +761,7 @@ generate_interaction_plot_configs <- function(df, limits_map = NULL) {
       x_breaks = levels(df_filtered$conc_num_factor),
       x_labels = levels(df_filtered$conc_num_factor),
       x_label = unique(df_filtered$Drug[1]),
-      coord_cartesian = y_range
+      coord_cartesian = y_range,
     )
 
     # Scatter plot config
@@ -768,7 +769,8 @@ generate_interaction_plot_configs <- function(df, limits_map = NULL) {
       plot_type = "scatter",
       title = sprintf("%s      %s", df_filtered$OrfRep[1], df_filtered$Gene[1]),
       error_bar = TRUE,
-      position = "jitter"
+      position = "jitter",
+      size = 1
     ))
 
     # Box plot config
@@ -937,7 +939,7 @@ generate_rank_plot_configs <- function(df, variables, is_lm = FALSE, adjust = FA
           )
         ),
         shape = 3,
-        size = 0.1,
+        size = 0.25,
         smooth = TRUE,
         smooth_color = "black",
         lm_line = list(intercept = intercept, slope = slope),

qhtcp-workflow/qhtcp-workflow

@@ -1438,6 +1438,10 @@ wrapper calculate_interaction_zscores
 # * Add gene names, other threshold values, etc.
 # * Dataframe columns and output file columns should be standardized in calculate_interactions()
 # * Need to decide if conc_num_factor is numeric or factor
+# * Do pdfs really need to be all different sizes?
+# * We are using standard error bars using the same se values as the data now (includes Bessel's correction)
+#    * Plate analysis error bars and some others will be slightly different
+#    * Can be changed back but better to have plots reflect data, no?
 #
 # INPUT
 #