library(GenomicRanges)
library(rtracklayer)
library(BSgenome)
library(BSgenome.Scerevisiae.UCSC.sacCer3)
library(tidyverse)

saccer3 <- BSgenome.Scerevisiae.UCSC.sacCer3

promoters <- rtracklayer::import("~/code/hf/yeast_genome_resources/yiming_promoters.bed")
promoters_unique <- promoters[!duplicated(paste(seqnames(promoters), start(promoters), end(promoters), strand(promoters)))]
promoter_seqs <- BSgenome::getSeq(saccer3, promoters_unique)

# Genome-wide base frequencies (all nuclear chromosomes, both strands balanced so +/- cancel)
genome_seqs <- BSgenome::getSeq(saccer3, seqnames(saccer3))
genome_freq <- Biostrings::alphabetFrequency(genome_seqs, as.prob = FALSE)
genome_freq_acgt <- colSums(genome_freq[, c("A", "C", "G", "T")])
genome_prop <- genome_freq_acgt / sum(genome_freq_acgt)

genome_df <- tibble(
    basepair = names(genome_prop),
    freq = genome_prop
)

# Promoter position frequency matrix
cm <- Biostrings::consensusMatrix(promoter_seqs)
pm <- prop.table(cm, margin = 2)
pm_acgt <- pm[c("A", "C", "G", "T"), ]

pm_acgt %>%
    as_tibble(rownames = "basepair") %>%
    pivot_longer(-basepair) %>%
    mutate(name = as.numeric(str_remove(name, "V"))) %>%
    filter(name <= 700) %>%
    left_join(genome_df, by = "basepair") %>%
    ggplot(aes(color = basepair)) +
    geom_line(aes(name, value)) +
    geom_hline(
        data = genome_df,
        aes(yintercept = freq, color = basepair),
        linetype = "dashed"
    ) +
    labs(x = "position", y = "proportion")