scripts/rossi_peak_analysis.R

library(tidyverse)
library(janitor)
library(GenomicRanges)
library(here)

exclude_regions <- rtracklayer::import(here("data/ChExMix_Peak_Filter_List_190612.bed"))
seqlevels(exclude_regions)[which(seqlevels(exclude_regions) == "chr2-micron")] <- "2-micron"
brentlab_features <- read_csv("~/projects/huggingface/yeast_genome_resources/brentlab_features.csv.gz")


promoters <- list(
    bp500 = rtracklayer::import("~/projects/huggingface/yeast_genome_resources/start_codon_500bp_upstream_promoters.bed"),
    mindel = GenomicRanges::GRanges(read_csv("~/projects/huggingface/yeast_genome_resources/mindel_promoters.csv.gz")),
    kang = rtracklayer::import("~/projects/huggingface/yeast_genome_resources/yiming_promoters.bed"),
    intergenic = rtracklayer::import("~/projects/huggingface/yeast_genome_resources/intergenic_regions_5_1.bed")
)

GenomicRanges::mcols(promoters$mindel) <- GenomicRanges::mcols(promoters$mindel) |>
    as.data.frame() |>
    dplyr::transmute(name = target_locus_tag) |>
    S4Vectors::DataFrame()

read_in_annotated_peaks <- function(peak_path) {
    df <- read_tsv(peak_path, show_col_types = FALSE)

    # Skip empty annotations
    if (nrow(df) == 0) {
        warning(sprintf("Skipping empty annotation file: %s", basename(peak_path)))
        return(NULL)
    }

    tryCatch(
        {
            # Convert peaks to GRanges for overlap detection
            peaks_gr <- GenomicRanges::GRanges(
                seqnames = df$Chr,
                ranges = IRanges::IRanges(start = df$Start, end = df$End),
                strand = df$Strand
            )

            # Find overlaps with exclude regions
            overlaps <- GenomicRanges::findOverlaps(peaks_gr, exclude_regions)

            # Label peaks that overlap with exclude regions
            df <- df |>
                mutate(
                    in_exclude_region = seq_len(nrow(df)) %in% S4Vectors::queryHits(overlaps)
                ) |>
                janitor::clean_names()

            return(df)
        },
        error = function(e) {
            warning(sprintf("Error processing: %s. Error: %s", basename(peak_path), e$message))
            return(NULL)
        }
    )
}

score_targets <- function(df, min_dist = 0, max_dist = 700) {
    df |>
        filter(!in_exclude_region) |>
        filter(
            peak_score > -log10(0.1),
            str_detect(nearest_promoter_id, "mRNA"),
            between(distance_to_tss, min_dist, max_dist)
        ) |>
        group_by(entrez_id) |>
        reframe(
            n_peaks = n(),
            nearest_score = peak_score[which.min(abs(distance_to_tss))],
            median_score = median(peak_score),
            max_score = max(peak_score)
        )
}


annotated_peaks <- list(
    files = list.files(here("data/chipexo_macs3/annotated_peaks"),
        "_annotated_peaks.txt",
        full.names = TRUE,
        recursive = TRUE
    )
)
names(annotated_peaks$files) <- str_remove(
    basename(annotated_peaks$files),
    "_annotated_peaks.txt"
)

annotated_peaks$df <- map(annotated_peaks$files, read_in_annotated_peaks)

annotated_peaks$dfcomp

annotated_peaks$target_score <- map(compact(annotated_peaks$df), score_targets)

peaks_df <- bind_rows(annotated_peaks$target_score, .id = "tmp") |>
    separate_wider_delim(tmp,
        delim = "_", names = c(
            "regulator_locus_tag",
            "regulator_symbol",
            "treatment",
            "growth_media"
        ),
        too_few = "align_start"
    ) |>
    mutate(
        treatment = ifelse(treatment == "Heat", "Heat Shock", treatment),
        growth_media = ifelse(is.na(growth_media), "YPD", growth_media)
    )


annotate_peaks_to_promoters <- function(peaks_df, promoters_gr) {
    peaks_gr <- GenomicRanges::GRanges(
        seqnames = peaks_df$chr,
        ranges = IRanges::IRanges(start = peaks_df$start, end = peaks_df$end)
    )

    hits <- GenomicRanges::findOverlaps(peaks_gr, promoters_gr, ignore.strand = TRUE)

    if (length(hits) == 0) {
        return(peaks_df |> dplyr::slice(0) |> dplyr::mutate(promoter_id = character(), distance_to_tss = numeric()))
    }

    promoter_strand <- as.character(GenomicRanges::strand(promoters_gr))
    promoter_start <- GenomicRanges::start(promoters_gr)
    promoter_end <- GenomicRanges::end(promoters_gr)
    promoter_name <- promoters_gr$name

    peaks_df[S4Vectors::queryHits(hits), ] |>
        dplyr::mutate(
            promoter_id = promoter_name[S4Vectors::subjectHits(hits)],
            .promoter_strand = promoter_strand[S4Vectors::subjectHits(hits)],
            .promoter_start = promoter_start[S4Vectors::subjectHits(hits)],
            .promoter_end = promoter_end[S4Vectors::subjectHits(hits)],
            .peak_mid = (start + end) / 2,
            # TSS-proximal edge of the promoter interval
            .tss_pos = dplyr::if_else(.promoter_strand == "+", .promoter_end, .promoter_start),
            distance_to_tss = abs(.peak_mid - .tss_pos)
        ) |>
        dplyr::select(-dplyr::starts_with("."), distance_to_tss, promoter_id)
}

score_targets_promoters <- function(df, promoters_gr, peak_score_thresh = -log10(0.1)) {
    df |>
        dplyr::filter(!in_exclude_region) |>
        annotate_peaks_to_promoters(promoters_gr) |>
        dplyr::filter(peak_score > peak_score_thresh) |>
        dplyr::group_by(promoter_id) |>
        dplyr::reframe(
            n_peaks = n(),
            nearest_score = peak_score[which.min(distance_to_tss)],
            median_score = median(peak_score),
            max_score = max(peak_score)
        )
}

chrmap <- read_csv("~/projects/huggingface/yeast_genome_resources/chrmap.csv.gz")

reduce_peak_cols <- function(df) {
    df |>
        dplyr::rename(peak_id = 1) |>
        dplyr::select(peak_id, chr, start, end, strand, peak_score, in_exclude_region) |>
        left_join(dplyr::select(chrmap, chr, ucsc)) |>
        mutate(chr = ucsc) |>
        dplyr::select(-ucsc)
}


annotated_peaks$df_reduced <- map(
    compact(annotated_peaks$df),
    reduce_peak_cols
)

annotated_peaks_by_promoter <- map(promoters, ~ {
    map(
        annotated_peaks$df_reduced,
        score_targets_promoters,
        promoters_gr = .
    )
})

intergenic_meta <- read_csv("~/projects/huggingface/yeast_genome_resources/intergenic_regions_metadata_5_1.csv")

annotated_peaks_by_promoter_df <- map(annotated_peaks_by_promoter,
    bind_rows,
    .id = "tmp"
)

annotated_peaks_by_promoter_df$intergenic <- annotated_peaks_by_promoter_df$intergenic |>
    left_join(intergenic_meta |>
        dplyr::select(
            promoter_id = ir_name,
            feature_left,
            feature_right
        ) |>
        pivot_longer(-promoter_id, values_to = "target_locus_tag") |>
        dplyr::select(-name), relationship = "many-to-many") |>
    dplyr::select(-promoter_id) |>
    mutate(promoter_id = target_locus_tag)

chec_meta <- arrow::read_parquet("~/projects/huggingface/rossi_2021/rossi_2021_metadata_sample.parquet")

reformat_tmp <- function(df) {
    df |>
        separate_wider_delim(tmp,
            delim = "_", names = c(
                "regulator_locus_tag",
                "regulator_symbol",
                "treatment",
                "growth_media"
            ),
            too_few = "align_start"
        ) |>
        mutate(
            treatment = ifelse(treatment == "Heat", "Heat Shock", treatment),
            growth_media = ifelse(is.na(growth_media), "YPD", growth_media)
        ) |>
        mutate(target_locus_tag = promoter_id) |>
        dplyr::select(-promoter_id) |>
        left_join(dplyr::select(brentlab_features,
            target_locus_tag = locus_tag,
            target_symbol = symbol
        )) |>
        dplyr::relocate(regulator_locus_tag, regulator_symbol, treatment, growth_media, target_locus_tag, target_symbol) |>
        group_by(regulator_locus_tag, treatment, growth_media) |>
        arrange(desc(max_score)) |>
        ungroup() |>
        filter(
            !is.na(target_locus_tag),
            !is.na(target_symbol)
        ) |>
        left_join(dplyr::select(
            chec_meta,
            sample_id, regulator_locus_tag,
            treatment, growth_media
        )) |>
        dplyr::relocate(sample_id)
}

annotated_peaks_by_promoter_df_out <- map(annotated_peaks_by_promoter_df, reformat_tmp)

write_out_promoter_intersect_peaks <- function(name, df) {
    output_path <- file.path(
        "~/projects/huggingface/rossi_2021",
        paste0("macs_", name, ".parquet")
    )
    df |>
        dplyr::select(-c(regulator_locus_tag, regulator_symbol, treatment, growth_media)) |>
        arrow::write_parquet(output_path)
}

# map2(names(annotated_peaks_by_promoter_df_out),
#      annotated_peaks_by_promoter_df_out,
#      write_out_promoter_intersect_peaks)

rossi_sample_meta <- arrow::read_parquet("~/projects/huggingface/rossi_2021/rossi_2021_metadata_sample.parquet")

brentlab_features <- read_csv("~/projects/huggingface/yeast_genome_resources/brentlab_features.csv.gz")

peaks_df_to_hf <- peaks_df |>
    left_join(rossi_sample_meta) |>
    dplyr::select(sample_id, regulator_locus_tag,
        regulator_symbol,
        target_locus_tag = entrez_id,
        n_peaks, nearest_score, median_score, max_score
    ) |>
    left_join(dplyr::select(brentlab_features,
        target_locus_tag = locus_tag,
        target_symbol = symbol
    )) |>
    dplyr::relocate(
        sample_id, regulator_locus_tag, regulator_symbol,
        target_locus_tag, target_symbol
    )

# peaks_df_to_hf |>
#     arrow::write_parquet("~/projects/huggingface/rossi_2021/macs2_annotated_peaks.parquet")

# mcisaac_responsive <- arrow::read_parquet("~/projects/huggingface/hackett_2020/hackett_2020_analysis_set.parquet")
#
# peaks_with_mcisaac <- peaks_df |>
#     dplyr::rename(target_locus_tag = entrez_id) |>
#     filter(
#         treatment == "Normal",
#         growth_media == "YPD",
#         median_score >= -log10(0.1)
#     ) |>
#     left_join(dplyr::select(
#         mcisaac_responsive,
#         regulator_locus_tag,
#         target_locus_tag,
#         time,
#         responsive
#     )) |>
#     filter(!is.na(responsive))
#
# peaks_with_mcisaac |>
#     filter(time == 30) |>
#     group_by(regulator_locus_tag) |>
#     nest() |>
#     mutate(
#         rr_nearest = map_dbl(data, ~ {
#             .x |>
#                 arrange(desc(nearest_score)) |>
#                 slice_head(n = 25) |>
#                 summarise(sum(responsive) / n()) |>
#                 pull()
#         }),
#         rr_max = map_dbl(data, ~ {
#             .x |>
#                 arrange(desc(max_score)) |>
#                 slice_head(n = 25) |>
#                 summarise(sum(responsive) / n()) |>
#                 pull()
#         }),
#         rr_median = map_dbl(data, ~ {
#             .x |>
#                 arrange(desc(median_score)) |>
#                 slice_head(n = 25) |>
#                 summarise(sum(responsive) / n()) |>
#                 pull()
#         })
#     ) |>
#     dplyr::select(-data) |>
#     pivot_longer(
#         cols = starts_with("rr_"),
#         names_to = "score_type",
#         values_to = "rr"
#     ) |>
#     ggplot(aes(x = score_type, y = rr)) +
#     geom_boxplot()
#
# library(patchwork)
# library(gridExtra)
#
# score_summary <- peaks_with_mcisaac |>
#     filter(time == 30) |>
#     group_by(regulator_locus_tag) |>
#     reframe(
#         score_type = c("nearest", "max", "median"),
#         n_targets = c(
#             n_distinct(target_locus_tag),
#             n_distinct(target_locus_tag),
#             n_distinct(target_locus_tag)
#         ),
#         n_peaks = c(n(), n(), n()),
#         min = c(min(nearest_score), min(max_score), min(median_score)),
#         max = c(max(nearest_score), max(max_score), max(median_score)),
#         median = c(median(nearest_score), median(max_score), median(median_score)),
#         mean = c(mean(nearest_score), mean(max_score), mean(median_score))
#     )
#
#
# p1 <- score_summary |>
#     ggplot(aes(x = score_type, y = mean, fill = score_type)) +
#     geom_boxplot(alpha = 0.7) +
#     labs(title = "Mean Score Distribution", y = "Mean Score", x = "") +
#     theme_minimal() +
#     theme(legend.position = "none")
#
# p3 <- score_summary |>
#     ggplot(aes(x = n_targets, y = mean, color = score_type)) +
#     geom_point(alpha = 0.6) +
#     facet_wrap(~score_type) +
#     scale_x_log10() +
#     labs(title = "Number of Targets vs Mean Score", x = "N Targets (log10)", y = "Mean") +
#     theme_minimal() +
#     theme(legend.position = "none")
#
# # Calculate stats for table
# n_targets_summary <- score_summary |>
#     dplyr::select(n_targets) |>
#     distinct() |>
#     pull(n_targets) %>%
#     {
#         tibble(
#             Min = round(quantile(., probs = 0), 2),
#             Q25 = round(quantile(., probs = 0.25), 2),
#             Median = round(quantile(., probs = 0.5), 2),
#             Q75 = round(quantile(., probs = 0.75), 2),
#             Max = round(quantile(., probs = 1), 2)
#         )
#     }
#
# # Vertical boxplot
# p4_plot <- score_summary |>
#     dplyr::select(regulator_locus_tag, n_targets) |>
#     distinct() |>
#     ggplot(aes(x = "", y = n_targets)) +
#     geom_boxplot(width = 0.3) +
#     scale_y_log10() +
#     labs(title = "Distribution of Targets per Regulator", y = "N Targets (log10)", x = "") +
#     theme_minimal() +
#     theme(legend.position = "none")
#
# # Summary table
# p4_table <- gridExtra::tableGrob(n_targets_summary,
#     rows = NULL,
#     theme = ttheme_minimal(base_size = 10)
# )
#
# (p1 + p3) / (p4_plot + p4_table)
#
# authors_orig_peaks <- arrow::read_parquet("~/projects/huggingface/rossi_2021/yep_filtered_peaks.parquet") |>
#     mutate(yeastepigenome_id = as.integer(yeastepigenome_id))
# authors_orig_peaks_meta <- arrow::read_parquet("~/projects/huggingface/rossi_2021/rossi_2021_metadata.parquet")
#
# authors_orig_peaks_normal_conds <- authors_orig_peaks |>
#     left_join(authors_orig_peaks_meta) |>
#     filter(treatment == "Normal", growth_media == "YPD")
#
# find_nearest_peaks <- function(macs_peaks, yep_chexmix_peaks, chrmap) {
#     library(GenomicRanges)
#     # Prepare MACS peaks (convert chr names)
#     macs_gr <- macs_peaks |>
#         left_join(chrmap |> dplyr::select(ucsc, chr)) |>
#         dplyr::select(-chr) |>
#         dplyr::rename(seqnames = ucsc) |>
#         filter(!is.na(seqnames)) |>
#         dplyr::select(seqnames, start, end, macs_score = peak_score) |>
#         GRanges()
#
#     # Prepare YEP ChExMix peaks (convert chr names)
#     yep_gr <- yep_chexmix_peaks |>
#         dplyr::select(seqnames = chr, start, end, yeastepigenome_id, yep_score = score) |>
#         GRanges()
#
#     # Find nearest neighbors
#     hits <- distanceToNearest(yep_gr, macs_gr)
#
#     # Add results back to YEP peaks
#     yep_with_nearest <- yep_chexmix_peaks |>
#         mutate(
#             query_idx = 1:n(),
#             subject_idx = subjectHits(hits),
#             distance = mcols(hits)$distance
#         ) |>
#         left_join(
#             macs_peaks |>
#                 mutate(subject_idx = 1:n()) |>
#                 dplyr::select(subject_idx, macs_score = peak_score, nearest_promoter_id),
#             by = "subject_idx"
#         ) |>
#         mutate(
#             macs_score_percentile = percent_rank(macs_score)
#         )
#
#     return(yep_with_nearest)
# }
#
# chrmap <- read_csv("~/projects/huggingface/yeast_genome_resources/chrmap.csv.gz")
#
# # Usage:
# authors_with_nearest <- find_nearest_peaks(
#     annotated_peaks$df$YJR060W_CBF1_Normal_YPD,
#     authors_orig_peaks_normal_conds |> filter(regulator_symbol == "CBF1"),
#     chrmap
# )
#
# norm_cond_reg_syms <- intersect(
#     str_extract(names(compact(annotated_peaks$df))[str_detect(names(compact(annotated_peaks$df)), "Normal")], "(?<=_)[^_]+(?=_)"),
#     unique(authors_orig_peaks_normal_conds$regulator_symbol)
# )
#
#
# results <- map(norm_cond_reg_syms, ~ {
#     yep_df <- filter(authors_orig_peaks_normal_conds, regulator_symbol == .x)
#     rlt <- unique(yep_df$regulator_locus_tag)
#     macs_df <- annotated_peaks$df[[paste(rlt, .x, "Normal_YPD", sep = "_")]] |>
#         filter(peak_score > -log10(0.05))
#
#     find_nearest_peaks(
#         macs_df,
#         yep_df,
#         chrmap
#     )
# })
#
# names(results) <- norm_cond_reg_syms
# results_df <- bind_rows(results)
#
# summary(results_df$macs_score_percentile)