scripts/parse_kemmeren_data.R

library(tidyverse)
library(here)
library(httr)
library(arrow)
library(readxl)

# genomic feature harmonization table ----
# see https://huggingface.co/datasets/BrentLab/yeast_genome_resources
gene_table = arrow::open_dataset(here("data/genome_files/hf/features")) %>%
  as_tibble()

## there is a mislabeling in both regulator and target
## with YDR022C. The common name is given as CIS1. However,
## SGD reports YDR022C as ATG31. CIS1 systematic ID is YLR346C
## That is labeled as ATG31.
## This appears to be a swap

## regulator and geneSymbol LUG1 is actually YCR087C-A, which was
## made an alias but only documented on SGD, not actually in the
## GFF/GTF. This needs to be updated in the gene_table as an alias

add_datatype_to_colnames = function(df, skip_indicies){
  # Suffixes to append
  suffixes <- c("_M", "_A", "_pval")

  # Repeat the suffixes to match the length of my_vector
  repeated_suffixes <- rep(suffixes, length.out = length(colnames(df)[-skip_indicies]))

  # Append the suffixes to each element of my_vector
  modified_vector <- paste0(colnames(df)[-skip_indicies], repeated_suffixes)

  colnames(df)[-skip_indicies] = modified_vector

  # drop the first row, which is the "data type" row in the original data
  # where the entries are M, A and P_value. These entries are added to the
  # colname
  df[-1,]
}

get_clean_headers = function(path, skip_indicies = 1:3) {
  headers <- read_tsv(path, n_max = 1, name_repair = "minimal") %>%
    add_datatype_to_colnames(skip_indicies = skip_indicies) %>%
    colnames()

  # Replace " vs" (optionally followed by ".") with ";"
  headers <- str_replace(headers, " vs\\.? ", ";")

  # Find empty (or NA) headers and replace with X1, X2, ...
  empties <- which(is.na(headers) | headers == "")
  if (length(empties) > 0) {
    headers[empties] <- paste0("X", seq_along(empties))
  }

  headers
}

read_in_kemmeren_data = function(path, ...){
  read.delim(path,
             sep='\t',
             skip=2,
             check.names=FALSE,
             col.names=get_clean_headers(path, ...)) %>%
    as_tibble()
}

stopifnot(identical(get_clean_headers(here('data/kemmeren/deleteome_all_mutants_ex_wt_var_controls.txt.xz')),
                    get_clean_headers(here('data/kemmeren/deleteome_all_mutants_controls.txt.xz'))))

deleteome_all_mutants_controls =
  read_in_kemmeren_data(here('data/kemmeren/deleteome_all_mutants_controls.txt.xz'))

deleteome_ex_wt_var_controls =
  read_in_kemmeren_data(here('data/kemmeren/deleteome_all_mutants_ex_wt_var_controls.txt.xz'))

by_hand_locustag_map = tibble(
  systematicName = c('YAR062W', 'YDL038C', 'snR10',     'YGR272C',   'YIL080W',    'YIL168W', 'YIR044C'),
  locus_tag      = c('YAR061W', 'YDL039C', 'YNCG0013W', 'YGR271C-A', 'YIL082W-A', 'YIL167W', 'YIR043C')) %>%
  deframe()

by_hand_symbol_map = gene_table %>%
  filter(locus_tag %in% by_hand_locustag_map) %>%
  select(locus_tag, symbol) %>%
  deframe()

# note that by using the target_locus_tag and target_symbol,
# the YCR087C-A,  YLR352W nomenclature is fixed (in original,
# YCR087C-A was called LUG1, but that name was removed in 2012 per SGD.)
# Additionally, the YDR022C/CIS1 error is corrected by using target_locus_tag
# and target_symbol, and aligns with the deletion evidence (the TF labelled
# ATG31/YDR022C is KOed at YDR022C, not YLR346C, which is what is
# currently labeled CIS1)
target_df = deleteome_all_mutants_controls %>%
  select(reporterId, systematicName, geneSymbol) %>%
  distinct() %>%
  left_join(select(gene_table, locus_tag, symbol) %>%
              mutate(systematicName = locus_tag)) %>%
  mutate(locus_tag = ifelse(is.na(locus_tag), by_hand_locustag_map[systematicName], locus_tag)) %>%
  mutate(symbol = ifelse(is.na(symbol), by_hand_symbol_map[locus_tag], symbol)) %>%
  group_by(locus_tag) %>%
  mutate(multiple_probes = n()>1) %>%
  ungroup() %>%
  mutate(variable_in_wt = reporterId %in%
           setdiff(deleteome_all_mutants_controls$reporterId,
                   deleteome_ex_wt_var_controls$reporterId)) %>%
  dplyr::rename(target_locus_tag = locus_tag,
                target_symbol = symbol)

rm(deleteome_ex_wt_var_controls)
gc()

deleteome_all_mutants_controls_long = deleteome_all_mutants_controls %>%
  pivot_longer(-c(reporterId, systematicName, geneSymbol),
               names_to='sample_metric', values_to='values') %>%
  separate(sample_metric, c('sample', 'metric'), sep="_") %>%
  pivot_wider(names_from='metric', values_from='values') %>%
  separate_wider_delim(cols=sample,
                       names=c('kemmeren_regulator', 'control'),
                       delim=";") %>%
  mutate(kemmeren_regulator = toupper(str_remove(tolower(kemmeren_regulator), "-del-1$|-del-mata$|-del$"))) %>%
  mutate(kemmeren_regulator = ifelse(kemmeren_regulator == "ARG5,6", "ARG56", kemmeren_regulator))

kem_sup1_regulator_info = read_excel(here("data/kemmeren/supplemental_table1_strain_info.xlsx")) %>%
  mutate(`profile first published` = str_replace(`profile first published`, ", ", ","))
kem_sup1_regulator_info_straininfo = read_excel(here("data/kemmeren/supplemental_table1_strain_info_origins.xlsx"))

kem_sup1_regulator_info = kem_sup1_regulator_info %>%
  left_join(kem_sup1_regulator_info_straininfo) %>%
  mutate(`profile first published` = citation) %>%
  select(-citation)

parsed_regulators = deleteome_all_mutants_controls_long %>%
  select(kemmeren_regulator) %>%
  distinct()

regulators_munging_list = list()

regulators_munging_list$x1 = parsed_regulators %>%
  mutate(gene = kemmeren_regulator) %>%
  left_join(kem_sup1_regulator_info) %>%
  filter(complete.cases(.))

regulators_munging_list$x2 = parsed_regulators %>%
  filter(!kemmeren_regulator %in% regulators_munging_list$x1$kemmeren_regulator) %>%
  mutate(`orf name` = kemmeren_regulator) %>%
  left_join(kem_sup1_regulator_info) %>%
  filter(complete.cases(.))

stopifnot(length(intersect(regulators_munging_list$x1$kemmeren_regulator, regulators_munging_list$x2)) == 0)

regulators_munging_list$x3 = read_csv(here("data/kemmeren/supplement_failure_regulator_mapping.csv.gz"))

stopifnot(length(intersect(regulators_munging_list$x2$kemmeren_regulator, regulators_munging_list$x3$kemmeren_regulator)) == 0)

regulators_munging_df = bind_rows(regulators_munging_list) %>%
  # the orf name for these two was the symbol
  mutate(`orf name` = case_when(
    `orf name` == "TLC1" ~ "YNCB0010W",
    `orf name` == "CMS1" ~ "YLR003C",
    .default = `orf name`
  )) %>%
  filter(kemmeren_regulator != "LUG1") %>%
  bind_rows(tibble(
    kemmeren_regulator = "LUG1",
    gene = "YCR087C-A",
    `orf name` = "YCR087C-A",
    description = paste0("Protein of unknown function; binds zinc; phosphomutants ",
                         "exhibit phenotypes, suggesting functionality of phosphosites; green ",
                         "fluorescent protein (GFP)-fusion protein localizes to the nucleolus; ",
                         "YCR087C-A is not an essential gene"),
    `functional category` = "unknown",
    `slide(s)` = "THM_00005835_S01 / THM_00005836_S01",
    `mating type` = "MATalpha",
    `source of deletion mutant(s)` = "Open Biosystems / Open Biosystems",
    `primary Hybset(s)` = "THM006 / THM006",
    `responsive/non-responsive` = "responsive mutant",
    chase_notes = paste0("This was originally called LUG1. However, that name ",
                         "for this locus was removed in 2012 per SGD. The expression confirms ",
                         "that the KO locus is YCR087C-A, not YLR352W, which is the locus ",
                         "currently called LUG1 in 2025"))) %>%
  left_join(
    select(gene_table, locus_tag, symbol) %>%
      mutate(`orf name` = locus_tag) %>%
      dplyr::rename(regulator_locus_tag = locus_tag,
                    regulator_symbol = symbol)) %>%
  replace_na(list(chase_notes = "none")) %>%
  mutate(regulator_locus_tag = ifelse(str_detect(kemmeren_regulator, "^WT-"), kemmeren_regulator, regulator_locus_tag),
         regulator_symbol = ifelse(str_detect(kemmeren_regulator, "^WT-"), kemmeren_regulator, regulator_symbol)) %>%
  janitor::clean_names()


stopifnot(setequal(regulators_munging_df$kemmeren_regulator,
                   unique(deleteome_all_mutants_controls_long$kemmeren_regulator)))

deleteome_all_mutants_svd_transforms =
  read_tsv(here("data/kemmeren/deleteome_all_mutants_svd_transformed.txt.xz"),
           name_repair = "minimal")

colnames(deleteome_all_mutants_svd_transforms)[1] = "systematicName"

colnames(deleteome_all_mutants_svd_transforms) =
  str_replace(colnames(deleteome_all_mutants_svd_transforms), "mf.alpha.1", "mf(alpha)1")
colnames(deleteome_all_mutants_svd_transforms) =
  str_replace(colnames(deleteome_all_mutants_svd_transforms), "mf.alpha.2", "mf(alpha)2")
colnames(deleteome_all_mutants_svd_transforms) =
  str_replace(colnames(deleteome_all_mutants_svd_transforms), "arg5.6", "arg56")

deleteome_all_mutants_svd_transforms_long = deleteome_all_mutants_svd_transforms %>%
  dplyr::rename(geneSymbol = commonName) %>%
  pivot_longer(-c(systematicName, geneSymbol),
               names_to = "condition",
               values_to = "Madj")  %>%
  separate_wider_delim(condition,
                       names = c("kemmeren_regulator", "tmp"),
                       delim = ".",
                       too_many = "merge") %>%
  mutate(kemmeren_regulator = toupper(kemmeren_regulator)) %>%
  mutate(
    # these regulators are missing appropriate suffixes
    kemmeren_regulator = recode(kemmeren_regulator,
                                "YIL014C" = "YIL014C-A",
                                "YOL086W" = "YOL086W-A",
                                "YDR034W" = "YDR034W-B",
                                "YAL044W" = "YAL044W-A"
    ),
    # these targets are incorrectly labeled with symbols rather than systematic IDs
    systematicName = recode(systematicName,
                            "ANR2" = "YKL047W",
                            "CMS1" = "YLR003C"
    )
  ) %>%
  # this is not in the other kemmeren data
  filter(systematicName != "Q0010")

stopifnot(length(setdiff(deleteome_all_mutants_svd_transforms_long$kemmeren_regulator,
                         regulators_munging_df$kemmeren_regulator)) == 0)

stopifnot(length(setdiff(deleteome_all_mutants_svd_transforms_long$systematicName,
                         target_df$systematicName)) == 0)

final_parsed_list = list(
    all = deleteome_all_mutants_controls_long %>%
      select(reporterId, kemmeren_regulator, M, A, pval) %>%
      left_join(select(target_df, reporterId, target_locus_tag,
                       target_symbol, multiple_probes, variable_in_wt)),

    slow_growth = deleteome_all_mutants_svd_transforms_long %>%
      select(kemmeren_regulator, systematicName, Madj) %>%
      # necessary to wrap in distinct to eliminate cases where there are two reporterId
      left_join(distinct(select(target_df, systematicName, target_locus_tag, target_symbol))) %>%
      select(-systematicName)
)

final_parsed_df = Reduce(left_join, final_parsed_list) %>%
  group_by(kemmeren_regulator) %>%
  # since the slow growth removed data identifies records by systematicName
  # and not reporterId, there is a many-to-many join and one reporterId is
  # duplicated to mulitple Madj. This removes those duplicates
  distinct(reporterId, .keep_all = TRUE) %>%
  ungroup() %>%
  left_join(select(regulators_munging_df,
                   -c(gene, `orf_name`))) %>%
  dplyr::rename(nr_sign_changes = nr_sign_changes_p_0_05_fc_1_7,
                primary_hybsets = primary_hybset_s,
                source_of_deletion_mutants = source_of_deletion_mutant_s,
                slides = slide_s,
                regulator_desc = description) %>%
  arrange(regulator_locus_tag)
  # select(regulator_locus_tag, regulator_symbol, reporterId,
  #        target_locus_tag, target_symbol, M, Madj, A, pval,
  #        variable_in_wt, multiple_probes)

db_kemmeren_meta = read_csv("data/kemmeren/db_kemmeren_meta_20251126.csv") %>%
  mutate(id = ifelse(regulator_locus_tag == 'YLR352W', 0, id)) %>%
  select(id, regulator_locus_tag) %>%
  distinct() %>%
  mutate(id = as.integer(id)) %>%
  dplyr::rename(db_id = id)

final_df_parsed_with_ids = final_parsed_df %>%
  left_join(db_kemmeren_meta) %>%
  replace_na(list(db_id = 0)) %>%
  arrange(regulator_locus_tag) %>%
  group_by(regulator_locus_tag) %>%
  mutate(sample_id = cur_group_id()) %>%
  relocate(sample_id, db_id,
           regulator_locus_tag, regulator_symbol,
           reporterId, target_locus_tag, target_symbol,
           M, Madj, A, pval,
           variable_in_wt, multiple_probes)

# note! verify before overwriting that the sample_id for the unique sample
# tuple is the same as it is in the current hackett_2020, or that any changes
# are intentional

# final_df_parsed_with_ids %>%
#   write_parquet("~/code/hf/kemmeren_2014/kemmeren_2014.parquet",
#                 compression = "zstd",
#                 chunk_size = 6181,
#                 write_statistics = TRUE,
#                 use_dictionary = c(
#                   regulator_locus_tag = TRUE,
#                   target_locus_tag = TRUE
#                 )
#   )