drug_targets.Rnw

<<setup_targets, include=FALSE>>=
library(dplyr)
b = import('base')
gdsc = import('data/gdsc')
util = import('./util_3')
config = import('../config')

resp = c("speed_matrix", "mutation") %>%
    lapply(util$load_fun) %>%
    bind_rows() %>%
    mutate(method = config$id2short(method))

# how many drugs have lower FDR for mutations and PROGENy
tab = resp %>%
    filter(adj.p < 0.1) %>%
    group_by(drug) %>%
    top_n(1, -adj.p) %>%
    select(drug, method)
tab = tab[!b$duplicated(tab$drug),]
table(tab$method)

targets = gdsc$drug$targets() %>%
    select(-id) %>%
    distinct() %>%
    inner_join(tab, by="drug") %>%
    select(-drug, -synonyms, -brand_name, -target)


do_test = function(col) {
    fisher.test(targets[,"method"], targets[,col], alternative="two.sided") %>%
        broom::tidy() %>%
        mutate(type = col)
}
tt = setdiff(colnames(targets), c("method", "chromatin_other"))
result = bind_rows(lapply(tt, do_test)) %>%
    filter(estimate != Inf) %>%
    mutate(adj.p = p.adjust(p.value, method="fdr"))
@