I list below the R packages required to reproduce the analyses.
## data wrangling
library(dplyr)
library(tidyr)
library(stringr)
library(purrr)
## plotting
library(ggplot2)
library(ComplexHeatmap)
library(cowplot)
library(grid)
library(RColorBrewer)
library(igraph)
library(magick)
## auxiliary functions
source("R/utils.R")
today_date <- "2025-04-28"
## today_date <- format(Sys.Date(), "%Y-%m-%d")
## set the seed, for fixing generation of ComplexHeatmaps (dendogram clustering)
set.seed(20)Correlation matrix computed with stats::cor.
Yet, we have to try other correlation methods on the continuous space, or consider other metrics if working on the discrete space. Unbalanced Wassertein distance, or relatives, for distinct input and output dimensions.
Some tutorials:
Mapping quantitative data to color, from Gehlenborg and Wong (2012):
Real-Life Examples with ComplexHeatmap:
Real-life examples with ggplot2
Real-Life examples with funkyheatmap CRAN package
## Read SummarizedExperiment object
se_normalised_replicate <- readRDS(paste0("./results/compounds/se_normalised_per_replicate_", today_date, ".rds"))
se_normalised_replicate_mat <- SummarizedExperiment::assay(se_normalised_replicate)
se_normalised_pheno_data_replicate <- SummarizedExperiment::colData(se_normalised_replicate)
se_normalised_pheno_data_replicate$log10Conc <- if_else(se_normalised_pheno_data_replicate$Concentrations==0,
0,
-log10(se_normalised_pheno_data_replicate$Concentrations))
## Compute global correlation score
cor_per_replicate <- cor(se_normalised_replicate_mat)
## Define clustering colours
col_heatmap_scale <- circlize::colorRamp2(c(min(cor_per_replicate), 0, 1),
c("blue", "white", "red"))
## Define custom colour Heatmap annotation
## num_batches <- length(unique(se_normalised_pheno_data_replicate$Batch_ID))
## num_compounds <- length(unique(se_normalised_pheno_data_replicate$Compound))
## num_moa <- length(unique(se_normalised_pheno_data_replicate$MoA))
##
## batch_col <- setNames(colorRampPalette(brewer.pal(8, "Dark2"))(num_batches),
## unique(se_normalised_pheno_data_replicate$Batch_ID))
## compound_col <- setNames(colorRampPalette(brewer.pal(11, "Spectral"))(num_compounds),
## unique(se_normalised_pheno_data_replicate$Compound))
## moa_col <- setNames(colorRampPalette(brewer.pal(12, "Set3"))(num_moa),
## unique(se_normalised_pheno_data_replicate$MoA))
## Heatmap annotation
top_ha <- ComplexHeatmap::HeatmapAnnotation(df = se_normalised_pheno_data_replicate |>
as.data.frame() |>
dplyr::select(Batch_ID, MoA, Compound))
## col = list(Batch_ID = batch_col,
## MoA = moa_col,
## Compound = compound_col))
row_ha <- ComplexHeatmap::rowAnnotation(Concentration =
ComplexHeatmap::anno_barplot(se_normalised_pheno_data_replicate$log10Conc),
Duration = paste0(se_normalised_pheno_data_replicate$Duration, "d"))
## 1) Generate HeatMap
heatmap_normalised_replicates <- ComplexHeatmap::Heatmap(cor_per_replicate,
col=col_heatmap_scale,
name = "CC for all Replicates",
show_row_names = TRUE,
show_column_names = FALSE,
row_dend_reorder = TRUE,
column_dend_reorder = TRUE,
## font size
row_names_gp = gpar(fontsize = 2),
column_names_gp = gpar(fontsize = 2),
column_names_rot = 45,
top_annotation = top_ha,
right_annotation = row_ha,
## matrix size
width = ncol(cor_per_replicate)*unit(1, "mm"),
height = nrow(cor_per_replicate)*unit(1, "mm"))
size_heatmap <- calc_ht_size(heatmap_normalised_replicates)
padding_heatmap <- unit(c(2, 15, 2, 2), "mm")
## 2) Save png and pdf outputs
png(paste0("figures/compounds/heatmap_normalised_replicates_", today_date,".png"),
res = 200,
width = size_heatmap$width, height = size_heatmap$height, units = c("in"))
heatmap_normalised_replicates |> ComplexHeatmap::draw(padding = padding_heatmap)
dev.off()
## png
## 2
knitr::include_graphics(paste0("figures/compounds/heatmap_normalised_replicates_", today_date,".png"),
auto_pdf = TRUE,
dpi = 100)
ggsave(paste0("figures/compounds/heatmap_normalised_replicates_", today_date,".pdf"),
heatmap_normalised_replicates |>
ComplexHeatmap::draw(padding = padding_heatmap) |>
grid::grid.grabExpr(),
dpi = 100,
width = size_heatmap$width, height = size_heatmap$height, units = c("in"))
## Read SummarizedExperiment object
se_normalised_compound_batch <- readRDS(paste0("./results/compounds/se_normalised_per_compound_by_batch_", today_date, ".rds"))
se_normalised_compound_batch_mat <- SummarizedExperiment::assay(se_normalised_compound_batch)
se_pheno_data_normalised_compound_batch <- SummarizedExperiment::colData(se_normalised_compound_batch)
## Compute global correlation score
cor_per_compound_by_batch <- cor(se_normalised_compound_batch_mat)
## Define clustering colours
col_heatmap_scale <- circlize::colorRamp2(c(min(cor_per_compound_by_batch), 0, 1),
c("blue", "white", "red"))
## Heatmap annotation
ha <- ComplexHeatmap::HeatmapAnnotation(df = se_pheno_data_normalised_compound_batch |>
as.data.frame() |>
dplyr::select(Pathway))
## Define Heatmap
heatmap_normalised_compound_batch <- ComplexHeatmap::Heatmap(cor_per_compound_by_batch,
col=col_heatmap_scale,
name = "CC for all Compound",
show_row_names = TRUE,
show_column_names = TRUE,
row_dend_reorder = TRUE,
column_dend_reorder = TRUE,
row_names_gp = gpar(fontsize = 6),
column_names_gp = gpar(fontsize = 3),
column_names_rot = 45,
top_annotation = ha,
## matrix size
width = ncol(cor_per_compound_by_batch)*unit(2, "mm"),
height = nrow(cor_per_compound_by_batch)*unit(2, "mm"))
size_heatmap <- calc_ht_size(heatmap_normalised_compound_batch)
padding_heatmap <- unit(c(2, 15, 2, 2), "mm")
## 2) Save png and pdf outputs
png(paste0("figures/compounds/heatmap_normalised_compound_by_batch_", today_date,".png"),
res = 200,
width = size_heatmap$width, height = size_heatmap$height, units = c("in"))
heatmap_normalised_compound_batch |> ComplexHeatmap::draw(padding = padding_heatmap)
dev.off()
## png
## 2
knitr::include_graphics(paste0("figures/compounds/heatmap_normalised_compound_by_batch_", today_date,".png"),
auto_pdf = TRUE,
dpi = 200)
ggsave(paste0("figures/compounds/heatmap_normalised_compound_by_batch_", today_date,".pdf"),
heatmap_normalised_compound_batch |>
ComplexHeatmap::draw(padding = padding_heatmap) |>
grid::grid.grabExpr(),
dpi = 200,
width = size_heatmap$width, height = size_heatmap$height, units = c("in"))
## Read SummarizedExperiment object
se_normalised_compound <- readRDS(paste0("./results/compounds/se_normalised_per_compound_", today_date, ".rds"))
se_normalised_compound_mat <- SummarizedExperiment::assay(se_normalised_compound)
se_normalised_pheno_data_compound <- SummarizedExperiment::colData(se_normalised_compound)
## Compute global correlation score
cor_per_compound <- cor(se_normalised_compound_mat)
## Define clustering colours
col_heatmap_scale <- circlize::colorRamp2(c(min(cor_per_compound), 0, 1),
c("blue", "white", "red"))
## Heatmap annotation
ha <- ComplexHeatmap::HeatmapAnnotation(df = se_normalised_pheno_data_compound |>
as.data.frame() |>
dplyr::select(Pathway))
## Define Heatmap
heatmap_normalised_compound <- ComplexHeatmap::Heatmap(cor_per_compound,
col=col_heatmap_scale,
name = "CC for all Compound",
show_row_names = TRUE,
show_column_names = TRUE,
row_dend_reorder = FALSE,
column_dend_reorder = TRUE,
## font size
row_names_gp = gpar(fontsize = 6),
column_names_gp = gpar(fontsize = 6),
column_names_rot = 45,
top_annotation = ha,
## matrix size
width = ncol(cor_per_compound)*unit(3, "mm"),
height = nrow(cor_per_compound)*unit(3, "mm"))
size_heatmap <- calc_ht_size(heatmap_normalised_compound)
padding_heatmap <- unit(c(2, 15, 2, 2), "mm")
## 2) Save png and pdf outputs
png(paste0("figures/compounds/heatmap_normalised_compound_", today_date,".png"),
res = 300,
width = size_heatmap$width, height = size_heatmap$height, units = c("in"))
heatmap_normalised_compound |> ComplexHeatmap::draw(padding = padding_heatmap)
dev.off()
## png
## 2
knitr::include_graphics(paste0("figures/compounds/heatmap_normalised_compound_", today_date,".png"),
auto_pdf = TRUE,
dpi = 300)
ggsave(paste0("figures/compounds/heatmap_normalised_compound_", today_date,".pdf"),
heatmap_normalised_compound |>
ComplexHeatmap::draw(padding = padding_heatmap) |>
grid::grid.grabExpr(),
dpi = 300,
width = size_heatmap$width, height = size_heatmap$height, units = c("in"))
## Read SummarizedExperiment object
se_binarised_replicates <- readRDS(paste0("./results/compounds/se_discretised_per_replicate_", today_date, ".rds"))
se_binarised_replicates_mat <- SummarizedExperiment::assay(se_binarised_replicates)
se_binarised_pheno_data_replicate <- SummarizedExperiment::colData(se_binarised_replicates)
## Compute global correlation score
cor_per_replicates <- cor(se_binarised_replicates_mat)
## Define clustering colours
col_heatmap_scale <- circlize::colorRamp2(c(min(cor_per_replicates), 0, 1),
c("blue", "white", "red"))
## Heatmap annotation
ha <- ComplexHeatmap::HeatmapAnnotation(df = se_binarised_pheno_data_replicate |>
as.data.frame() |>
dplyr::select(Pathway, Batch_ID, Compound))
heatmap_binarised_replicates <- ComplexHeatmap::Heatmap(cor_per_replicates,
col=col_heatmap_scale,
name = "CC at binarised Compound Level",
show_row_names = TRUE,
show_column_names = TRUE,
row_dend_reorder = FALSE,
column_dend_reorder = TRUE,
## font size
row_names_gp = gpar(fontsize = 2),
column_names_gp = gpar(fontsize = 2),
column_names_rot = 45,
top_annotation = ha,
## matrix size
width = ncol(cor_per_replicate)*unit(1, "mm"),
height = nrow(cor_per_replicate)*unit(1, "mm"))
size_heatmap <- calc_ht_size(heatmap_binarised_replicates)
padding_heatmap <- unit(c(2, 15, 2, 2), "mm")
## 2) Save png and pdf outputs
png(paste0("figures/compounds/heatmap_binarised_replicates_", today_date,".png"),
res = 100,
width = size_heatmap$width, height = size_heatmap$height, units = c("in"))
heatmap_binarised_replicates |> ComplexHeatmap::draw(padding = padding_heatmap)
dev.off()
## png
## 2
knitr::include_graphics(paste0("figures/compounds/heatmap_binarised_replicates_", today_date,".png"),
auto_pdf = TRUE,
dpi = 100)
ggsave(paste0("figures/compounds/heatmap_binarised_replicates_", today_date,".pdf"),
heatmap_binarised_replicates |>
ComplexHeatmap::draw(padding = padding_heatmap) |>
grid::grid.grabExpr(),
dpi = 100,
width = size_heatmap$width, height = size_heatmap$height, units = c("in"))
## Read SummarizedExperiment object
se_binarised_compound_batch <- readRDS(paste0("./results/compounds/se_discretised_per_compound_by_batch_", today_date, ".rds"))
se_binarised_compound_batch_mat <- SummarizedExperiment::assay(se_binarised_compound_batch)
se_binarised_pheno_data_compound_batch <- SummarizedExperiment::colData(se_binarised_compound_batch)
## Compute global correlation score
cor_per_compound_by_batch <- cor(se_binarised_compound_batch_mat)
## Define clustering colours
col_heatmap_scale <- circlize::colorRamp2(c(min(cor_per_compound_by_batch), 0, 1),
c("blue", "white", "red"))
## Heatmap annotation
ha <- ComplexHeatmap::HeatmapAnnotation(df = se_binarised_pheno_data_compound_batch |>
as.data.frame() |>
dplyr::select(Pathway, MoA))
heatmap_binarised_compound_batch <- ComplexHeatmap::Heatmap(cor_per_compound_by_batch,
col=col_heatmap_scale,
name = "CC at binarised Compound Level",
show_row_names = TRUE,
show_column_names = TRUE,
row_dend_reorder = FALSE,
column_dend_reorder = TRUE,
## font size
row_names_gp = gpar(fontsize = 6),
column_names_gp = gpar(fontsize = 3),
column_names_rot = 45,
top_annotation = ha,
## matrix size
width = ncol(cor_per_compound_by_batch)*unit(2, "mm"),
height = nrow(cor_per_compound_by_batch)*unit(2, "mm"))
size_heatmap <- calc_ht_size(heatmap_binarised_compound_batch)
padding_heatmap <- unit(c(2, 15, 2, 2), "mm")
## 2) Save png and pdf outputs
png(paste0("figures/compounds/heatmap_binarised_compound_by_batch_", today_date,".png"),
res = 200,
width = size_heatmap$width, height = size_heatmap$height, units = c("in"))
heatmap_binarised_compound_batch |> ComplexHeatmap::draw(padding = padding_heatmap)
dev.off()
## png
## 2
knitr::include_graphics(paste0("figures/compounds/heatmap_binarised_compound_by_batch_", today_date,".png"),
auto_pdf = TRUE,
dpi = 200)
ggsave(paste0("figures/compounds/heatmap_binarised_compound_by_batch_", today_date,".pdf"),
heatmap_binarised_compound_batch |>
ComplexHeatmap::draw(padding = padding_heatmap) |>
grid::grid.grabExpr(),
dpi = 200,
width = size_heatmap$width, height = size_heatmap$height, units = c("in")) 
## Read SummarizedExperiment object
se_binarised_compound <- readRDS(paste0("./results/compounds/se_discretised_per_compound_", today_date, ".rds"))
se_binarised_compound_mat <- SummarizedExperiment::assay(se_binarised_compound)
se_binarised_pheno_data_compound <- SummarizedExperiment::colData(se_binarised_compound)
## Compute global correlation score
cor_per_compound <- cor(se_binarised_compound_mat)
## Define clustering colours
col_heatmap_scale <- circlize::colorRamp2(c(min(cor_per_compound), 0, 1),
c("blue", "white", "red"))
## Heatmap annotation
ha <- ComplexHeatmap::HeatmapAnnotation(df = se_binarised_pheno_data_compound |>
as.data.frame() |>
dplyr::select(Pathway, MoA))
heatmap_binarised_compound <- ComplexHeatmap::Heatmap(cor_per_compound,
col=col_heatmap_scale,
name = "CC at binarised Compound Level",
show_row_names = TRUE,
show_column_names = TRUE,
row_dend_reorder = TRUE,
column_dend_reorder = TRUE,
row_names_gp = gpar(fontsize = 8),
column_names_gp = gpar(fontsize = 8),
column_names_rot = 45,
top_annotation = ha,
## matrix size
width = ncol(cor_per_compound)*unit(4, "mm"),
height = nrow(cor_per_compound)*unit(4, "mm"))
size_heatmap <- calc_ht_size(heatmap_binarised_compound)
padding_heatmap <- unit(c(2, 10, 2, 2), "mm")
## 2) Save png and pdf outputs
png(paste0("figures/compounds/heatmap_binarised_compound_", today_date,".png"),
res = 300,
width = size_heatmap$width, height = size_heatmap$height, units = c("in"))
heatmap_binarised_compound |> ComplexHeatmap::draw(padding = padding_heatmap)
dev.off()
## png
## 2
knitr::include_graphics(paste0("figures/compounds/heatmap_binarised_compound_", today_date,".png"),
auto_pdf = TRUE,
dpi = 300)
ggsave(paste0("figures/compounds/heatmap_binarised_compound_", today_date,".pdf"),
heatmap_binarised_compound |>
ComplexHeatmap::draw(padding = padding_heatmap) |>
grid::grid.grabExpr(),
dpi = 300,
width = size_heatmap$width, height = size_heatmap$height, units = c("in")) 
igraph::graph_from_adjacency_matrix in Figure 3.7, filtering out pairwise connections between 2 drugs if the \(p\)-value is not significant and absolute correlation score is below 0.8:#| label: fig-igraph-drug
## 1) Compute correlation matrix and p-values ----
cor_res <- Hmisc::rcorr(se_binarised_compound_mat, type = "pearson")
cor_binarised_compound <- cor_res$r
pval_binarised_compound <- cor_res$P
adj_mat <- cor_binarised_compound
adj_mat[pval_binarised_compound > 0.05/ncol(cor_res)] <- 0
## 2) build igraph adjacency matrix ----
compound_graph <- igraph::graph_from_adjacency_matrix(adjmatrix=adj_mat,
weighted = TRUE,
diag = FALSE,
mode = "undirected")
## 3) igraph customisation ----
compound_graph$date <- today_date
E(compound_graph)$color <- if_else(E(compound_graph)$weight >0, "red", "blue")
E(compound_graph)$weight <- abs(E(compound_graph)$weight)
layout_graph <- igraph::layout_with_fr(compound_graph,
niter = 500)
# graph dimension sizes
graph_width <- range(layout_graph[,1]) |> diff()
graph_height <- range(layout_graph[,2]) |> diff()
aspect_ratio <- graph_width / graph_height
vertex_structure <- se_binarised_pheno_data_compound |>
as.data.frame() |>
dplyr::mutate(color = if_else(Pathway=="EGFR", "green", "yellow"))
igraph::V(compound_graph)$color <- vertex_structure$color
pdf(paste0("figures/compounds/compound_graph_", today_date, ".pdf"),
width = 14 * aspect_ratio, height = 14,
useDingbats = TRUE, compress = FALSE)
plot(compound_graph, layout = layout_graph, vertex.size = 10,
## vertex.label.dist= 2,
main = "Compound network interaction.
In green, drugs related with the EGFR mechanism,
and in yellow other pathways.
Edge widths are proportional to the coefficient of correlation.",
edge.width = igraph::E(compound_graph)$weight * 20,
vertex.label.cex = 0.9, curved = 0.2)
dev.off()
## png
## 2
## 4) export PNG and readable graph formats for Neo4J
png(paste0("figures/compounds/compound_graph_", today_date, ".png"),
width = 14 * aspect_ratio, height = 14,
res = 200, units = c("in"))
plot(compound_graph, layout = layout_graph, vertex.size = 10,
## vertex.label.dist= 2,
main = "Compound network interaction.
In green, drugs related with the EGFR mechanism,
and in yellow other pathways.
Edge widths are proportional to the coefficient of correlation.",
edge.width = igraph::E(compound_graph)$weight * 20,
vertex.label.cex = 0.9, curved = 0.2)
dev.off()
## png
## 2
knitr::include_graphics(paste0("figures/compounds/compound_graph_", today_date, ".png"),
dpi = 200, auto_pdf = TRUE)
igraph::write_graph(graph = compound_graph,
file = paste0("./results/compounds/compound_network_", today_date, ".graphml"),
format = "graphml")
se_binarised_compound_batch_fac <- apply(se_binarised_compound_batch_mat, 2, as.character)
## Heatmap annotation
ha <- ComplexHeatmap::HeatmapAnnotation(df = se_binarised_pheno_data_compound_batch |>
as.data.frame() |>
dplyr::select(Batch_ID, Pathway, Compound))
heatmap_cell_lines_by_sample <- ComplexHeatmap::Heatmap(se_binarised_compound_batch_fac,
col = c("TRUE" = "red", "FALSE" = "blue"),
name = "Heatmap: cell-line by sample",
show_row_names = FALSE,
row_dend_reorder = TRUE,
## clustering_distance_rows = "pearson",
show_row_dend = FALSE,
column_dend_reorder = TRUE,
show_column_names = TRUE,
column_names_gp = gpar(fontsize = 6),
column_names_rot = 45,
top_annotation = ha,
## matrix size
width = ncol(se_binarised_compound_batch_fac)*unit(3, "mm"),
height = nrow(se_binarised_compound_batch_fac)*unit(0.3, "mm"),
use_raster = TRUE)
size_heatmap <- calc_ht_size(heatmap_cell_lines_by_sample)
padding_heatmap <- unit(c(2, 10, 2, 2), "mm")
## 2) Save png and pdf outputs
png(paste0("figures/cell_lines/heatmap_cell_lines_by_samples_", today_date,".png"),
res = 50,
width = size_heatmap$width, height = size_heatmap$height, units = c("in"))
heatmap_cell_lines_by_sample |> ComplexHeatmap::draw(padding = padding_heatmap)
dev.off()
## png
## 2
knitr::include_graphics(paste0("figures/cell_lines/heatmap_cell_lines_by_samples_", today_date,".png"),
auto_pdf = TRUE,
dpi = 50)
ggsave(paste0("figures/cell_lines/heatmap_cell_lines_by_samples_", today_date,".pdf"),
heatmap_cell_lines_by_sample |>
ComplexHeatmap::draw(padding = padding_heatmap) |>
grid::grid.grabExpr(),
dpi = 200, limitsize = FALSE,
width = size_heatmap$width, height = size_heatmap$height, units = c("in")) 