Utility function to read and parse GMT file provided by Pathway Commons into R

read.pc2.gmt.R

library(stringr)
read.pc2.gmt <- function(file, organism=9606, datasource='kegg', idtype='hgnc symbol') {
    # utility function to read and parse GMT file
    # provided by Pathway Commons into R
    # i.e. https://www.pathwaycommons.org/archives/PC2/v12/PathwayCommons12.All.hgnc.gmt.gz
    gmt <- list()
    con <- file(file, open='r')
    for (line in readLines(con)) {
        line_parts <- unlist(strsplit(line, '\t'))
        desc_keys <- c('name', 'datasource', 'organism', 'idtype')
        indices <- str_locate(line_parts[2], paste0(desc_keys, ': '))
        desc_list <- list()
        for (i in seq_along(desc_keys)) {
            start <- indices[i,2]+1
            stop <- ifelse(i == length(desc_keys), nchar(line_parts[2]), indices[i+1,1]-3)
            desc_list[[ desc_keys[i] ]] <- substr(line_parts[2], start, stop)
        }
        genes <- line_parts[3:length(line_parts)]
        if (desc_list$organism==9606 && desc_list$datasource==datasource && desc_list$idtype==idtype) {
            gmt[[ desc_list$name ]] <- genes
        }
    }
    close(con)
    return(gmt)
}