Skip to content

Latest commit

 

History

History
80 lines (63 loc) · 3.22 KB

find_screen_klaeger_matches.md

File metadata and controls

80 lines (63 loc) · 3.22 KB

Find Screen-Klaeger Drug Matches

Matthew Berginski 2021-07-07

klaeger = read_rds(here('results/klaeger_full_tidy.rds')) %>%
    mutate(drug = trimws(drug))

screen = read_rds(here('results/screen_combined.rds')) %>%
    mutate(anchor = trimws(anchor), compound = trimws(compound))

klaeger_drugs = unique(klaeger$drug) %>% sort()
screen_drugs = unique(c(screen$anchor,screen$compound)) %>% sort()

perfect_matches = klaeger_drugs[klaeger_drugs %in% screen_drugs]

#remove the direct perfect matches from the above lists and then dump each list
#out to CSV for manual matching

klaeger_drugs_no_match = klaeger_drugs[! klaeger_drugs %in% perfect_matches]
screen_drugs_no_match = screen_drugs[! screen_drugs %in% perfect_matches]

write_csv(as.data.frame(klaeger_drugs_no_match), here('src/find_screen_klaeger_matches/no_clear_match_klaeger_drug_list.csv'))
write_csv(as.data.frame(screen_drugs_no_match), here('src/find_screen_klaeger_matches/no_clear_match_screen_drug_list.csv'))

From here I opened each individual drug list file and manually matched the drug names, producing “manual_match_list.csv”. Now I’ll open this file and run a few checks on the manual results and then combine them with the perfect match list from above.

manual_matches = read_csv(here('src/find_screen_klaeger_matches/manual_match_list.csv'))
## 
## ── Column specification ────────────────────────────────────────────────────────
## cols(
##   klaeger_drugs = col_character(),
##   screen_drugs = col_character()
## )

assertthat::assert_that(all(manual_matches$screen_drugs %in% screen_drugs_no_match))
## [1] TRUE

#Dinaciclib is annotated in two ways in the screen screen, as "Dinaciclib",
#which has a perfect match in the Klaeger set and as "Dinaciclib (SCH727965)"
#which doesn't have a perfect match in the Klaeger set. These are clearly the
#same drug though, so I added "Dinaciclib" to the manual match list, but since
#this is a perfect match in the Klaeger set it was removed from the no match
#list. This causes the assert to fail though, so I've filtered the Dinaciclib to
#"Dinaciclib (SCH727965)" match out for this check.
assertthat::assert_that(all(manual_matches %>% 
                                                            filter(klaeger_drugs != "Dinaciclib") %>% 
                                                            pull(klaeger_drugs) %in% klaeger_drugs_no_match))
## [1] TRUE

all_matches = bind_rows(
    manual_matches,
    data.frame(klaeger_drugs = perfect_matches,
                         screen_drugs = perfect_matches)
) %>% arrange(screen_drugs) %>%
    write_csv(here('src/find_screen_klaeger_matches/klaeger_screen_drug_matches.csv'))

unmatched_klaeger = data.frame(
    klaeger_drugs = klaeger_drugs[!klaeger_drugs %in% all_matches$klaeger_drugs])
unmatched_screen = data.frame(
    screen_drugs = screen_drugs[!screen_drugs %in% all_matches$screen_drugs])

write_csv(unmatched_klaeger,here('src/find_screen_klaeger_matches/unmatched_klaeger.csv'))
write_csv(unmatched_screen,here('src/find_screen_klaeger_matches/unmatched_screen.csv'))