work/multi-res-colors.R

In twesleyb/Uezu2016: iBioID Proteomics from Uezu, Kanak, Bradshaw et al., 2016

#!/usr/bin/env Rscript

# title: 
# description: generate module colors
# authors: Tyler W Bradshaw

## ---- INPUT

# input data in root/data
root = "~/projects/Uezu2016"

part_file = "cpm_partition"

# not clustered == "gray"
NC_color = "#BEBEBE" 


## ---- FUNCTIONS 

str_to_vec <- function(response) {
  # parse the python dictionary returned as a string from 
  # system(random_color.py)
	vec <- gsub("'","",gsub("\\[|\\]","",
				trimws(unlist(strsplit(response,",")))))
	return(vec)
}


## ---- Prepare the workspace 

# Global Imports
suppressPackageStartupMessages({
	library(dplyr)
	library(data.table)
})

# Local Imports 
devtools::load_all(root, quiet=TRUE)

# input multi-resolution partition
myfile <- file.path(root, "rdata", paste0(part_file,".csv"))
stopifnot(file.exists(myfile))

# load and coerce to list of vectors (partitions)
part_df <- fread(myfile,drop=1)
part_list <- unlist(apply(part_df,1,list), recursive=FALSE)

## we really only need to generate colors for the partition with max clusters
k = sapply(part_list,function(x) length(unique(x)))
r = which(k==max(k))

partition = part_list[[r]]

## ---- Generate Module colors 

prots <- names(partition)

modules <- split(prots, partition)
names(modules) <- paste0("M",names(modules))

n_colors <- length(modules) 
message("generating: ", n_colors, " random colors")

# Path to python script which is a simple script that uses the python 
# port of randomcolors to generate random colors.
script <- file.path(root,"Py","random_color.py")

stopifnot(file.exists(script))

# FIXME: how can we conveniently package this python code for R?

# Generate n random colors
cmd <- paste(script,"--count", n_colors, "--luminosity", "bright")
response <- system(cmd, intern = TRUE)


# NOTE: to install randomcolor, use conda:
# conda install -c conda-forge randomcolor

#  Parse the response
colors <- toupper(str_to_vec(response))

# Module color assignments
# Initialize a vector for the module colors
module_colors <- setNames(rep(NA, length(modules)), nm=names(modules))

# Insure that M0 is gray 
module_colors["M0"] <- NC_color

# The reamining colors are random
idx <- is.na(module_colors)
module_colors[idx] <- sample(colors,sum(idx))


## ---- work

module_colors

prots = names(partition)

modules = split(prots,partition)
names(modules) <- paste0("M", names(modules))

as.list(module_colors[names(modules)])

prot_colors = unlist(lapply(names(modules), function(x) setNames(rep(module_colors[[x]],length(modules[[x]])), nm=modules[[x]])))

part0 = part_list[[5]]
table(part0)


df = data.table(resolution=seq(1,100),k=k)
library(ggplot2)
plot <- ggplot(df)
plot <- plot + geom_line(aes(x=resolution,y=k))
plot <- plot + theme(panel.background=element_blank())
plot <- plot + theme(axis.line.x=element_line()) 
plot <- plot + theme(axis.line.y=element_line()) 
ggsave("cpm.pdf",plot)


## ---- Save the data 

# Save updated module colors
namen <- paste0(gsub("partition","colors",part_file),".rda")
myfile <- file.path(root,"data", namen)
save(module_colors,file=myfile,version=2)
message("saved: ", myfile)