Add figure generation script to .circleCI

- convert final png figure to pdf
- add figure generation script to `run-figures.sh`
- use color palette generated in PR AlexsLemonade#510 for figure color scheme
- format treemap to have less redundant values (treemap as is in final figure 1 panel does not show redundant values and represents the `short_histology` and `integrated_diagnosis` values which I believe should be fine in this case)
- redundant text does show up on the treemap plot in `analyses/sample-distribution-analysis/plots` directory (still looking into this)

.circleci/config.yml

@@ -20,6 +20,10 @@ jobs:
           name: Sample Distribution Analyses
           command: ./scripts/run_in_ci.sh bash "analyses/sample-distribution-analysis/run-sample-distribution.sh"
 
+      - run:
+          name: Sample Distribution Figure
+          command: ./scripts/run_in_ci.sh Rscript figures/scripts/fig1-sample-distribution.R
+
       - run:
           name: Transcriptome dimensionality reduction
           command: ./scripts/run_in_ci.sh ./analyses/transcriptomic-dimension-reduction/ci-dimension-reduction-plots.sh 

analyses/sample-distribution-analysis/02-multilayer-plots.R

@@ -67,50 +67,66 @@ final_df <- histologies_df %>%
 readr::write_tsv(final_df, file.path(results_dir, "plots_df.tsv"))
 
 # Create and save treemap using ggplot2
-# Number of palettes needed
-n <- length(unique(final_df$level3))
-
-# Now calculate the colors for each data point
+# Read in the histology color palette
+color_palette <-
+  readr::read_tsv(file.path(
+    root_dir,
+    "figures",
+    "palettes",
+    "histology_color_palette.tsv"
+  ))
+
+# Join the color palette for the colors for each short histology value --
+# palette is generated in `figures/scripts/color_palettes.R`
 final_df2 <- final_df %>%
-  dplyr::mutate(index = as.numeric(factor(final_df$level3))- 1) %>%
-  dplyr::group_by(index) %>%
-  dplyr::mutate(
-    max_size = max(size),
-    color = gradient_n_pal(
-      sequential_hcl(
-        6,
-        h = 360 * index[1]/n,
-        c = c(45, 20),
-        l = c(30, 80),
-        power = .5)
-    )(size/max_size)
-  )
-
-# Now plot using `geom_treemap` and save
-ggsave(
-ggplot(final_df2, aes(area = size, fill = color, label=level3, subgroup=level2, subgroup2=level1)) +
+  dplyr::left_join(color_palette, by = c("level2" = "color_names"))
+
+# Plot the treemap
+treemap <-
+  ggplot(
+    final_df2,
+    aes(
+      area = size,
+      fill = hex_codes,
+      label = level1,
+      subgroup = level2,
+      subgroup2 = level3
+    )
+  ) +
   geom_treemap() +
-  geom_treemap_subgroup_border(colour="white") +
-  geom_treemap_text(fontface = "italic",
-                    colour = "white",
-                    place = "centre",
-                    grow = F,
-                    reflow=T) +
-  geom_treemap_subgroup_text(place = "top",
-                             grow = T,
-                             reflow = T,
-                             alpha = 0.6,
-                             colour = "#FAFAFA",
-                             min.size = 0) +
-  geom_treemap_subgroup2_text(place = "centre",
-                             grow = T,
-                             alpha = 0.8,
-                             colour = "#FAFAFA",
-                             min.size = 0) +
-  scale_fill_identity(),
-file = file.path(plots_dir, "distribution_across_cancer_types_treemap.pdf"),
-width = 22,
-height = 10
+  geom_treemap_subgroup_border(colour = "white") +
+  geom_treemap_text(
+    fontface = "italic",
+    colour = "white",
+    place = "center",
+    alpha = 0.4,
+    grow = T,
+    reflow = T
+  ) +
+  geom_treemap_subgroup_text(
+    place = "top",
+    grow = T,
+    reflow = T,
+    alpha = 0.6,
+    colour = "#FAFAFA",
+    min.size = 0
+  ) +
+  geom_treemap_subgroup2_text(
+    place = "bottom",
+    grow = T,
+    reflow = T,
+    alpha = 0.5,
+    colour = "#FAFAFA",
+    min.size = 0
+  ) +
+  theme(legend.position = "none")
+
+# Save treemap
+ggsave(
+  treemap,
+  file = file.path(plots_dir, "distribution_across_cancer_types_treemap.pdf"),
+  width = 22,
+  height = 10
 )
 
 # Create a treemap (for interactive treemap)

figures/palettes/histology_color_palette.tsv

@@ -0,0 +1,38 @@
+color_names	hex_codes
+Adenoma	#f23d3d
+ATRT	#731d1d
+Central neurocytoma	#b38686
+Chondrosarcoma	#cc5c33
+Chordoma	#331c0d
+Choroid plexus tumor	#ffb380
+CNS EFT-CIC	#b25f00
+CNS lymphoma	#f2d6b6
+CNS neuroblastoma	#736556
+CNS Rhabdomyosarcoma	#ffaa00
+CNS sarcoma	#4c3d00
+Craniopharyngioma	#e2f200
+DNET	#919926
+Dysplasia	#d6f2b6
+Embryonal Tumor	#304d26
+Ependymoma	#00f241
+ETMR	#009929
+Ganglioglioma	#698c7c
+Germinoma	#39e6c3
+Glial-neuronal tumor NOS	#005359
+Gliosis	#263233
+Hemangioblastoma	#00c2f2
+Hemangioma	#40a6ff
+HGAT	#406280
+Langerhans Cell histiocytosis	#0044ff
+LGAT	#00144d
+LGMT	#acbbe6
+Medulloblastoma	#7373e6
+Meningioma	#3d0099
+MPNST	#c200f2
+Neurofibroma	#917399
+none	#f1f1f1
+Oligodendroglioma	#f279da
+Other	#cc0052
+Pineoblastoma	#994d6b
+Schwannoma	#4d2636
+Teratoma	#ffbfd9

figures/run-figures.sh

@@ -8,7 +8,12 @@ cd "$(dirname "${BASH_SOURCE[0]}")"
 # Make an output folder for all directories
 mkdir -p pngs
 
-## Sample distribution
+################ Sample distribution figure
+# Run sample distribution analysis
+bash ../analyses/sample-distribution-analysis/run-sample-distribution.sh
+
+# Run the figure assembly
+Rscript scripts/fig1-sample-distribution.R
 
 ################ Mutational landscape figure
 # Run both SNV caller consensus scripts

figures/scripts/fig1-sample-distribution.R

@@ -7,9 +7,11 @@
 
 # Load in libraries
 library(dplyr)
+library(ggplot2)
 library(colorspace)
 library(scales)
 library(treemapify)
+library(patchwork)
 
 # Magrittr pipe
 `%>%` <- dplyr::`%>%`
@@ -38,74 +40,72 @@ sample_distribution_dir <- file.path(
 disease_expression <-
   readr::read_tsv(file.path(sample_distribution_dir, "results", "disease_expression.tsv"))
 
-# Read in plots data.frame file from `02-multilaye-plots.R`
+# Read in plots data.frame file from `02-multilayer-plots.R`
 plots_df <-
   readr::read_tsv(file.path(sample_distribution_dir, "results", "plots_df.tsv"))
 
 # Reorder the columns to be displayed in descending order by count on the plot
 disease_expression$integrated_diagnosis <- with(disease_expression,
                                                 reorder(integrated_diagnosis, -count))
 
-#### Re-run the individual plots ----------------------------------------------
+# Read in the histology color palette
+color_palette <-
+  readr::read_tsv(file.path(
+    root_dir,
+    "figures",
+    "palettes",
+    "histology_color_palette.tsv"
+  ))
 
-# Create a bar plot of sample distribution across cancer types
-gg_types_bar <- disease_expression %>%
-  ggplot2::ggplot(ggplot2::aes(x = integrated_diagnosis, y = count, fill = count)) +
-  ggplot2::geom_col() +
-  ggplot2::theme_bw() +
-  ggplot2::labs(x = "Cancer Types", y = "Count",
-                title = "Sample Distribution Across Cancer Types") +
-  ggplot2::scale_y_continuous(breaks = seq(0, 500, by = 100)) +
-  ggplot2::theme(axis.text.x = ggplot2::element_text(
-    angle = 75,
-    hjust = 1,
-    size = 8
-  ),
-  panel.grid = ggplot2::element_blank()) +
-  ggplot2::geom_text(nudge_y = 5.5, size = 1,
-                     ggplot2::aes(label = paste0(disease_expression$percent)))
+#### Re-run the individual plots ----------------------------------------------
 
 # Create a treemap of broad histology, short histology, and integrated diagnosis
-# Number of palettes needed
-n <- length(unique(plots_df$level3))
 
-# Now calculate the colors for each data point
+# Join the color palette for the colors for each short histology value --
+# palette is generated in `figures/scripts/color_palettes.R`
 plots_df2 <- plots_df %>%
-  mutate(index = as.numeric(factor(plots_df$level3))- 1) %>%
-  group_by(index) %>%
-  mutate(
-    max_size = max(size),
-    color = gradient_n_pal(
-      sequential_hcl(
-        6,
-        h = 360 * index[1]/n,
-        c = c(45, 20),
-        l = c(30, 80),
-        power = .5)
-    )(size/max_size)
-  )
-
-# Plot the treemap
-treemap <- ggplot(plots_df2, aes(area = size, fill = color, label=level3, subgroup=level2, subgroup2=level1)) +
-    geom_treemap() +
-    geom_treemap_subgroup_border(colour="white") +
-    geom_treemap_text(fontface = "italic",
-                      colour = "white",
-                      place = "centre",
-                      grow = F,
-                      reflow=T) +
-    geom_treemap_subgroup_text(place = "top",
-                               grow = T,
-                               reflow = T,
-                               alpha = 0.6,
-                               colour = "#FAFAFA",
-                               min.size = 0) +
-    geom_treemap_subgroup2_text(place = "centre",
-                                grow = T,
-                                alpha = 0.8,
-                                colour = "#FAFAFA",
-                                min.size = 0) +
-    scale_fill_identity()
+  left_join(color_palette, by = c("level2" = "color_names"))
+
+# Plot the treemap where level1 is `broad_histology`,
+# level2 is `short_histology`, and level3 is `integrated_diagnosis`
+treemap <-
+  ggplot(
+    plots_df2,
+    aes(
+      area = size,
+      fill = hex_codes,
+      label = level1,
+      subgroup = level2,
+      subgroup2 = level3
+    )
+  ) +
+  geom_treemap() +
+  geom_treemap_subgroup_border(colour = "white") +
+  geom_treemap_text(
+    fontface = "italic",
+    colour = "white",
+    place = "center",
+    alpha = 0.4,
+    grow = T,
+    reflow = T
+  ) +
+  geom_treemap_subgroup_text(
+    place = "top",
+    grow = T,
+    reflow = T,
+    alpha = 0.6,
+    colour = "#FAFAFA",
+    min.size = 0
+  ) +
+  geom_treemap_subgroup2_text(
+    place = "bottom",
+    grow = T,
+    reflow = T,
+    alpha = 0.5,
+    colour = "#FAFAFA",
+    min.size = 0
+  ) +
+  theme(legend.position = "none")
 
 ## TODO: Re-run Github Contributions plot/table here -- for now we will define
 ## this plot as NULL
@@ -117,26 +117,19 @@ project_assays_plot <- NULL
 project_features_plot <- NULL
 
 #### Assemble multipanel plot -------------------------------------------------
-ggpubr::ggarrange(ggpubr::ggarrange(treemap,
-                                    project_features_plot,
-                                    ncol = 2,
-                                    labels = c("A", "C"),
-                                    widths = c(2, 1.6),
-                                    font.label = list(size = 20)
-),
-ggpubr::ggarrange(project_assays_plot,
-                  github_contributions_plot,
-                  ncol = 2,
-                  labels = c("B", "D"),
-                  widths = c(1.6, 2),
-                  font.label = list(size = 20)
-),
-nrow = 2,
-heights = c(2, 2)
-)
 
-# Save to PNG
-ggplot2::ggsave(file.path(output_dir, "fig1-openpbta-distribution.png"),
-                width = 16.5, height = 15,
+# Combine plots with patchwork
+# Layout of the four plots will be two over the other two
+# (2 columns and 2 rows)
+combined_plot <- treemap + project_features_plot +
+  project_assays_plot + github_contributions_plot +
+  plot_layout(ncol = 2, nrow = 2) &
+  theme(# add uniform labels
+    axis.text.x = element_text(size = 9),
+    axis.text.y = element_text(size = 9))
+
+# Save to PDF
+ggplot2::ggsave(file.path(output_dir, "fig1-openpbta-distribution.pdf"),
+                width = 10, height = 8,
                 units = "in"
 )