scripts/parse_hu_reiman_data.R

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

## NOTES:
##
## Hu had a heat shock and tetracycline condition. Reimand says they
## use:
## Further analysis was performed separately for the following
## three cases: A) same-strain replicates (BY4741);
## B) replicates of different strains (BY4741 and S288C);
## c) replicates of the induced promoter strain (R1158+TET)
##
## which i interpreted to mean (given some inspection from the original Hu
## metadata), that it should be !heat_shock and !tetracycline except
## for the regulators which did not have those conditions, in which
## case I kept the tetracyclie condition.

# 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()


hu_data = list(
  de = here("data/hu/gkq232_SuppTable1C_KOTargetGenes_Matrix_LogFoldChange.dat.xz"),
  pval = here("data/hu/gkq232_SuppTable1B_KOTargetGenes_Matrix_PValue.dat.xz")
)

hu_de_target_genes = as.character(read_delim(hu_data$de,
                             delim=' ',
                             col_names = FALSE,
                             n_max = 1)[1,])
hu_de_target_genes = str_replace(hu_de_target_genes, "\\.", "-")

hu_de_colnames = c('hu_regulator', hu_de_target_genes)

hu_pval_target_genes = as.character(read_delim(hu_data$pval,
                                             delim=' ',
                                             col_names = FALSE,
                                             n_max = 1)[1,])
hu_pval_target_genes = str_replace(hu_pval_target_genes, "\\.", "-")

hu_pval_colnames = c('hu_regulator', hu_pval_target_genes)

stopifnot(setequal(hu_pval_colnames, hu_de_colnames))

# Note: The warning is expected. this is being parsed
# correctly despite the warning
de_df = read.delim(
  hu_data$de,
  sep=' ',
  col.names = hu_de_colnames,
  check.names=FALSE) %>%
  as_tibble()

# Note: The warning is expected. this is being parsed
# correctly despite the warning
pval_df = read.delim(
  hu_data$pval,
  sep=' ',
  col.names = hu_pval_colnames,
  check.names=FALSE) %>%
  as_tibble()

hu_df = de_df %>%
  pivot_longer(-hu_regulator,
               names_to='hu_target',
               values_to='effect') %>%
  left_join(pval_df %>%
              pivot_longer(-hu_regulator,
                           names_to='hu_target',
                           values_to='pval')) %>%
  mutate(hu_regulator = toupper(hu_regulator)) %>%
  # remove deleted orfs
  filter(!hu_target %in% c('YER187W-A', 'YCR103C', 'YCL006C', "YIL080W"))

tf_map = list()

tf_map$init = hu_df %>%
  distinct(hu_regulator) %>%
  mutate(regulator_symbol = hu_regulator) %>%
  # these are labeled with aliases in the hu data
  mutate(regulator_symbol = case_when(
    regulator_symbol == "CAF17" ~ "IBA57",
    regulator_symbol == "RCS1" ~ "AFT1",
    regulator_symbol == "RLR1" ~ "THO2",
    regulator_symbol == "ZMS1" ~ "RSF2",
    regulator_symbol == "RIS1" ~ "ULS1",
    .default = regulator_symbol
  )) %>%
  left_join(gene_table %>%
              select(symbol,locus_tag) %>%
              dplyr::rename(regulator_locus_tag = locus_tag,
                            regulator_symbol = symbol)) %>%
  mutate(regulator_symbol = ifelse(is.na(regulator_locus_tag),
                                   NA,
                                   regulator_symbol))

tf_map$locus_tags = tf_map$init %>%
  filter(!complete.cases(.)) %>%
  select(hu_regulator) %>%
  mutate(regulator_locus_tag = hu_regulator) %>%
  left_join(gene_table %>%
              select(symbol,locus_tag) %>%
              dplyr::rename(regulator_locus_tag = locus_tag,
                            regulator_symbol = symbol))

tf_map_df = bind_rows(
  tf_map$init %>% filter(complete.cases(.)),
  tf_map$locus_tags
)

stopifnot(setequal(tf_map$init$hu_regulator, tf_map_df$hu_regulator))

target_map = list()

target_map$init = hu_df %>%
  distinct(hu_target) %>%
  mutate(target_locus_tag = hu_target) %>%
  left_join(select(gene_table, locus_tag, symbol)
            %>% dplyr::rename(target_locus_tag = locus_tag,
                              target_symbol = symbol))

target_map$alias_df = gene_table %>%
  mutate(hu_target = map_chr(alias, ~ {
    alias_i <- .x
    match <- keep(target_map$init %>%
                    filter(!complete.cases(.)) %>%
                    pull(hu_target),
                  ~ str_detect(alias_i, .x))
    if (length(match) > 0) match[[1]] else NA_character_
  })) %>%
  filter(!is.na(hu_target)) %>%
  select(hu_target, locus_tag, symbol) %>%
  dplyr::rename(target_locus_tag = locus_tag,
                target_symbol = symbol)

target_map_df = bind_rows(target_map$init %>% filter(complete.cases(.)),
                          target_map$alias_df)

stopifnot(setequal(target_map_df$hu_target,
                   unique(hu_df$hu_target)))

final_df = hu_df %>%
  left_join(tf_map_df) %>%
  left_join(target_map_df)

# never used -- don't bother
# hu_reimand_from_db = read_csv("data/hu/hu_2007_reimand_db_20251127.csv")

# from https://www.nature.com/articles/ng2012#Sec10
supplement_table_1 = readxl::read_excel("data/hu/41588_2007_BFng2012_MOESM20_ESM.xlsx") %>%
  janitor::clean_names() %>%
  dplyr::rename(regulator_locus_tag = `tf_systematic_name`) %>%
  select(-tf_gene_name) %>%
  mutate(heat_shock = str_detect(growth_condition, "heat-shock"),
         tetracycline_treatment = str_detect(growth_condition, "tetracycline treatment")) %>%
  select(-growth_condition)

supplement_table_1_filt = supplement_table_1 %>%
  filter(!heat_shock, !tetracycline_treatment)

supplement_table_1_metadata = supplement_table_1_filt %>%
  bind_rows(
    supplement_table_1 %>%
      filter(
        !heat_shock,
        regulator_locus_tag %in%
               setdiff(final_df$regulator_locus_tag,
                       supplement_table_1_filt$regulator_locus_tag)))

final_df_for_parquet = final_df %>%
  mutate(sample_id = cur_group_id()) %>%
  select(sample_id,
         regulator_locus_tag, regulator_symbol,
         target_locus_tag, target_symbol,
         effect, pval) %>%
  ungroup() %>%
  arrange(regulator_locus_tag) %>%
  left_join(supplement_table_1_metadata)


final_df_for_parquet %>%
  write_parquet(here("~/code/hf/hu_2007_reimand_2010/hu_2007_reimand_2010.parquet"),
                compression = "zstd",
                chunk_size = 6249,
                write_statistics = TRUE,
                use_dictionary = c(
                  sample_id = TRUE,
                  regulator_locus_tag = TRUE,
                  regulator_symbol = TRUE,
                  target_locus_tag = TRUE,
                  target_symbol = TRUE,
                  heat_shock = TRUE,
                  tetracycline_treatment = TRUE
                ))