scripts/parse_mahendrawada_rnaseq.R

library(tidyverse)
library(GEOquery)

count_parent_dir = "~/htcf_lts/downloaded_data/mahendrawada_2024/GSE236948_RAW"
counts_files = list.files(count_parent_dir, ".txt.gz")
names(counts_files) = str_extract(counts_files, "GSM\\d+")

genomic_features = arrow::read_parquet("~/code/hf/yeast_genome_resources/brentlab_features.parquet")

test_count_df = read_tsv(file.path(count_parent_dir, counts_files[[1]]), col_names = c('orig_locus_tag', 'count'))

find_matching_locus <- function(target, genomic_df) {
    match_idx <- which(str_detect(genomic_df$alias, fixed(target)))

    if (length(match_idx) == 0) {
        return(tibble(locus_tag = NA_character_, chr = NA_character_,
                      start = NA_real_, end = NA_real_, strand = NA_character_))
    } else if (length(match_idx) > 1) {
        # Handle multiple matches - take first or warn
        message(paste("Multiple matches for", target, "- taking first"))
        match_idx <- match_idx[1]
    }

    genomic_df %>%
        slice(match_idx) %>%
        select(locus_tag, chr, start, end, strand)
}

alias_matched_locus_tags <- test_count_df %>%
    mutate(target_locus_tag = case_when(
        orig_locus_tag == "LSR1"  ~ "YNCB0019C",
        orig_locus_tag == "SCR1"  ~ "YNCE0024W",
        orig_locus_tag == "snR18" ~ "YNCA0003W",
        orig_locus_tag == "snR19" ~ "YNCN0005C",
        orig_locus_tag == "snR3"  ~ "YNCJ0030W",
        orig_locus_tag == "snR39" ~ "YNCG0014C",
        orig_locus_tag == "snR4"  ~ "YNCE0019W",
        orig_locus_tag == "snR5"  ~ "YNCO0028W",
        orig_locus_tag == "snR6"  ~ "YNCL0006W",
        orig_locus_tag == "snR8"  ~ "YNCO0026W",
        .default = orig_locus_tag)) %>%
    left_join(genomic_features,
              by = c('target_locus_tag' = 'locus_tag')) %>%
    filter(is.na(chr), str_detect(target_locus_tag, "^__", negate=TRUE)) %>%
    select(orig_locus_tag, target_locus_tag) %>%
    mutate(target_locus_tag = str_replace_all(target_locus_tag, "\\(|\\)", "_")) %>%
    mutate(match_info = map(target_locus_tag, ~find_matching_locus(.x, genomic_features))) %>%
    unnest(match_info) %>%
    select(orig_locus_tag, locus_tag)

locus_tag_map = test_count_df %>%
    mutate(locus_tag = case_when(
        orig_locus_tag == "LSR1"  ~ "YNCB0019C",
        orig_locus_tag == "SCR1"  ~ "YNCE0024W",
        orig_locus_tag == "snR18" ~ "YNCA0003W",
        orig_locus_tag == "snR19" ~ "YNCN0005C",
        orig_locus_tag == "snR3"  ~ "YNCJ0030W",
        orig_locus_tag == "snR39" ~ "YNCG0014C",
        orig_locus_tag == "snR4"  ~ "YNCE0019W",
        orig_locus_tag == "snR5"  ~ "YNCO0028W",
        orig_locus_tag == "snR6"  ~ "YNCL0006W",
        orig_locus_tag == "snR8"  ~ "YNCO0026W",
        .default = orig_locus_tag)) %>%
    filter(locus_tag %in% genomic_features$locus_tag) %>%
    select(-count) %>%
    bind_rows(alias_matched_locus_tags) %>%
    left_join(select(genomic_features, locus_tag, symbol)) %>%
    dplyr::rename(target_locus_tag = locus_tag,
                  target_symbol = symbol)

gse_soft <- getGEO("GSE236947", GSEMatrix = FALSE)

# Extract all sample metadata including protocols
samples <- lapply(GSMList(gse_soft), function(gsm) {
    meta <- Meta(gsm)

    data.frame(
        gsm_id = meta$geo_accession,
        title = meta$title,
        sra = {
            rel <- meta$relation
            sra_rel <- rel[grep("SRA", rel)]
            if(length(sra_rel) > 0) sub(".*term=", "", sra_rel[1]) else NA
        },
        genotype = ifelse(!is.null(meta$characteristics_ch1),
                          {
                              # Extract genotype from characteristics
                              char_lines <- meta$characteristics_ch1
                              genotype_line <- char_lines[grep("genotype:", char_lines)]
                              if(length(genotype_line) > 0) {
                                  sub("genotype: ", "", genotype_line[1])
                              } else {
                                  NA
                              }
                          },
                          NA),
        growth_protocol = ifelse(!is.null(meta$growth_protocol),
                                 paste(meta$growth_protocol, collapse = " "),
                                 NA),
        treatment_protocol = ifelse(!is.null(meta$treatment_protocol),
                                    paste(meta$treatment_protocol, collapse = " "),
                                    NA),
        stringsAsFactors = FALSE
    )
})

metadata <- do.call(rbind, samples) %>%
    as_tibble() %>%
    dplyr::rename(sra_accession = sra) %>%
    # NOTE: this is a typo. It is correct in the
    # genotpye. There is no FHK2
    mutate(title = str_replace(title, "FHK2", "FKH2")) %>%
    # only GSM8316241, 2, 3 have Gal4 rather than GAL4 in
    # the title. The genotypes are exactly the same as the
    # other mnase construct RNA-seq samples. This appears
    # to be a typo, so it is being corrected
    mutate(title = str_replace(title, "Gal4", "GAL4"))


stopifnot(nrow(metadata) == length(counts_files))

metadata_parsed <- metadata %>%
    mutate(
        timepoint = str_extract(title, "_(10|30|60|120)_"),
        timepoint = as.integer(str_remove_all(timepoint, "_")),
        # Parse treatment_protocol into degron vs non-degron
        has_degron_system = str_detect(treatment_protocol, "DMSO|IAA"),

        # Extract degron treatment (DMSO vs IAA)
        degron_treatment = case_when(
            str_detect(title, "_DMSO_") ~ "DMSO",
            str_detect(title, "_3IAA_") ~ "IAA",
            TRUE ~ NA_character_
        ),

        # Detect any IAA7-based degron tag
        has_degron_tag = str_detect(genotype, "IAA7|mini-N-deg|mini-degron"),

        # Extract degron variant
        degron_variant = case_when(
            str_detect(genotype, "mini-N-deg") ~ "mini_N_terminal_IAA7",
            str_detect(genotype, "IAA7") ~ "full_or_short_IAA7",
            TRUE ~ NA_character_
        ),

        # Check for AID system components
        has_ostir1 = str_detect(genotype, "pGPD1-OSTIR"),

        # Parse genotype components
        # Extract TF-MNase fusion if present
        tf_mnase_fusion = str_extract(genotype, "[A-Z0-9]+(?=-MNase)"),

        # Extract regulator_symbol/TF name from title
        regulator_symbol = str_extract(title, "^[A-Z0-9]+(?=_)"),

        # Determine strain type
        strain_type = case_when(
            !is.na(regulator_symbol) ~ "degron_experiment",
            !is.na(tf_mnase_fusion) ~ "mnase_fusion",
            TRUE ~ "parent_strain"
        ),

        # Parse environmental condition from title
        env_condition = case_when(
            str_detect(title, "_SM_") ~ "SM",
            str_detect(title, "_no_gal_") ~ "no_galactose",  # Add this first
            str_detect(title, "_gal_") ~ "galactose",
            str_detect(title, "_raf_") ~ "raffinose",
            str_detect(title, "_37") | str_detect(title, "heat") ~ "heat_shock_37C",
            TRUE ~ "standard_30C"
        ),

        # Determine if this is WT control
        is_wt_control = str_detect(title, "^WT_"),

        # Extract replicate
        replicate = str_extract(title, "(?<=_)[ABC](?=_)"),

        # Classify experiment type
        experiment_type = case_when(
            has_degron_system & degron_treatment %in% c("DMSO", "IAA") & regulator_symbol == "WT" ~ "wt_degron_control",
            has_degron_system & degron_treatment %in% c("DMSO", "IAA") & regulator_symbol != "WT" ~ "degron",
            !has_degron_system & !is.na(tf_mnase_fusion) ~ "mnase_fusion_rnaseq",
            !has_degron_system & is_wt_control ~ "wt_baseline",
            TRUE ~ "other"
        ))

metadata_parsed_grouped = metadata_parsed %>%
    group_by(experiment_type)

metadata_parsed_split = metadata_parsed_grouped %>%
    group_split(.keep=FALSE)
names(metadata_parsed_split) = group_keys(metadata_parsed_grouped)$experiment_type

# for the degron samples, if the timepoint is not specified, it is 30 minutes.
# see mahnedrawada 2025 paper
metadata_parsed_split$degron = metadata_parsed_split$degron %>%
    replace_na(list(timepoint = 30))

get_counts_add_meta = function(gsmid){
    df = read_tsv(
        file.path(count_parent_dir,
                  counts_files[[gsmid]]),
        col_names = c('orig_locus_tag', 'count'))

    counts_meta = df %>%
        filter(str_detect(orig_locus_tag, "^__")) %>%
        mutate(orig_locus_tag = str_remove(orig_locus_tag, "__")) %>%
        dplyr::rename(tmp = orig_locus_tag) %>%
        pivot_wider(names_from = tmp, values_from = count) %>%
        mutate(gsm_id = gsmid) %>%
        dplyr::relocate(gsm_id)

    counts_df = df %>%
        filter(str_detect(orig_locus_tag, "^__", negate=TRUE)) %>%
        left_join(locus_tag_map) %>%
        mutate(gsm_id = gsmid) %>%
        select(gsm_id, target_locus_tag, target_symbol, orig_locus_tag, count)

    list(
        meta = counts_meta,
        counts = counts_df
    )

}

counts_parsed_list = map(metadata_parsed_split, ~{
    temp_list = map(.x$gsm_id, get_counts_add_meta)

    list(
        meta = map_df(temp_list, "meta") %>%
            left_join(select(.x, gsm_id, sra_accession), by = "gsm_id") %>%
            dplyr::relocate(sra_accession, gsm_id),
        counts = map_df(temp_list, "counts") %>%
            left_join(select(.x, gsm_id, sra_accession), by = "gsm_id") %>%
            dplyr::relocate(sra_accession) %>%
            dplyr::select(-gsm_id)
    )
})

counts_common_cols = c('sra_accession', 'gsm_id', 'replicate',
                       'no_feature', 'ambiguous', 'too_low_aQual',
                       'alignment_not_unique')
final_metadata_columns = list(
    degron = c('degron_treatment', 'degron_variant',
               'regulator_symbol', 'env_condition',
               'timepoint'),
    mnase_fusion_rnaseq = c('sra_accession', 'gsm_id', 'regulator_symbol', 'env_condition'),
    wt_baseline = c('env_condition'),
    wt_degron_control = c('degron_treatment')
)

metadata_with_counts_meta = map(names(metadata_parsed_split), ~{
    metadata_parsed_split[[.x]] %>%
        left_join(counts_parsed_list[[.x]]$meta) %>%
        select(all_of(c(final_metadata_columns[[.x]], counts_common_cols))) %>%
        dplyr::relocate(sra_accession, gsm_id) %>%
        dplyr::rename(gsm_accession = gsm_id)
})
names(metadata_with_counts_meta) = names(metadata_parsed_split)

set_names_grouping_col_list = list(
    wt_baseline = 'env_condition',
    wt_degron_control = 'degron_treatment',
    degron = c('degron_treatment', 'env_condition',
               'regulator_symbol', 'timepoint'),
    mnase_fusion_rnaseq = c('regulator_symbol', 'env_condition'))

prev_largest_sample_id = 0
metadata_with_counts_meta_with_sample_id = metadata_with_counts_meta
for(n in names(set_names_grouping_col_list)){
    metadata_with_counts_meta_with_sample_id[[n]] = metadata_with_counts_meta[[n]] %>%
        group_by(!!!syms(set_names_grouping_col_list[[n]])) %>%
        mutate(sample_id = cur_group_id() + prev_largest_sample_id) %>%
        arrange(sample_id)
    prev_largest_sample_id = max(metadata_with_counts_meta_with_sample_id[[n]]$sample_id)

}

metadata_with_counts_meta_with_sample_id$mnase_fusion_rnaseq =
    metadata_with_counts_meta_with_sample_id$mnase_fusion_rnaseq %>%
    left_join(select(genomic_features, symbol, locus_tag) %>%
                  dplyr::rename(regulator_symbol = symbol,
                                regulator_locus_tag = locus_tag)) %>%
    dplyr::relocate(sra_accession, gsm_accession,
                    regulator_locus_tag, regulator_symbol)

metadata_with_counts_meta_with_sample_id$degron =
    metadata_with_counts_meta_with_sample_id$degron %>%
    mutate(regulator_symbol = str_replace(regulator_symbol, "YNR063W", "PUL4")) %>%
    left_join(select(genomic_features, symbol, locus_tag) %>%
                  dplyr::rename(regulator_symbol = symbol,
                                regulator_locus_tag = locus_tag)) %>%
    dplyr::relocate(sra_accession, gsm_accession,
                    regulator_locus_tag, regulator_symbol)

for(n in names(metadata_with_counts_meta_with_sample_id)){
    meta = list(
        path = file.path("/home/chase/code/hf/mahendrawada_2025",
                             paste0(n, "_counts_meta.parquet")),
        data = metadata_with_counts_meta_with_sample_id[[n]]
    )

    counts = list(
        path = file.path("/home/chase/code/hf/mahendrawada_2025",
                             paste0(n, "_counts.parquet")),
        data = counts_parsed_list[[n]]$counts
    )

    arrow::write_parquet(
        meta$data,
        meta$path,
        compression = "zstd",
        write_statistics = TRUE)

    arrow::write_parquet(
        counts$data,
        counts$path,
        chunk_size = 7036,
        compression = "zstd",
        write_statistics = TRUE,
        use_dictionary = c(
            sra_accession = TRUE))
}