R/tss_filter.R

In primirTSS: Prediction of pri-miRNA Transcription Start Site

tss_single_judge2 <- function(stem_loop_p1, stem_loop_p2, chrom, strand, tss_p1,
                              tss_p2, gene_loci) {

  gene_df <- gene_loci[gene_loci$chrom == chrom & gene_loci$strand == strand, ]

  intra_index <- stem_loop_p1 >= gene_df$gene_p1 & stem_loop_p2 <=
    gene_df$gene_p2
  if (any(intra_index)) {
    gene_intra <- gene_df[intra_index, ]
    closest_tss <- lapply(gene_intra$gene_start, function(x) {
      min(abs(x - tss_p1), abs(x - tss_p2))}) %>%
      unlist()
    in_tss <- (gene_intra$gene_start - tss_p1) *
      (gene_intra$gene_start - tss_p2) < 0
    colsest_index <- closest_tss <= 150 | in_tss
    if (any(colsest_index)) {
      result <- paste("intra__host_TSS",
                      gene_intra$gene_id[colsest_index],
                      gene_intra$gene_start[colsest_index],
                      # closest_tss[colsest_index],
                      sep = "__") %>%
        paste0(collapse = ":")
    } else {
      tss_site <- ifelse(strand == "+", tss_p2, tss_p1)
      both_site_index <- (tss_site - gene_intra$gene_p1) *
        (tss_site - gene_intra$gene_p2) < 0
      if (any(both_site_index)) {
        gene_intra_2 <- gene_intra[both_site_index, ]
        distance <- abs(tss_site - gene_intra_2$gene_start)
        gene_intra_2_index <- which.min(distance)
        result <- paste("intra__intra_TSS",
                        gene_intra_2$gene_id[gene_intra_2_index],
                        tss_site,
                        sep = "__")
      } else {
        distance <- abs(tss_site - gene_intra$gene_start)
        gene_intra_2_index <- which.min(distance)
        result <- paste("intra__intra_TSS2",
                        gene_intra$gene_id[gene_intra_2_index],
                        tss_site,
                        sep = "__")
      }
    }
  } else {
    closest_tss <- lapply(gene_df$gene_start, function(x) {
      min(abs(x - tss_p1), abs(x - tss_p2))
    }) %>%
      unlist()
    in_tss <- (gene_df$gene_start - tss_p1) * (gene_df$gene_start - tss_p2) < 0
    colsest_index <- closest_tss <= 150 | in_tss

    if (any(colsest_index)) {
      tss_site <- ifelse(strand == "+", tss_p2, tss_p1)
      mir_end <- ifelse(strand == "+", stem_loop_p2, stem_loop_p1)

      distance <- abs(tss_site - gene_df$gene_start)
      gene_inter_index <- which.min(distance)
      result <- paste("inter__overlap_inter_TSS2",
                      gene_df$gene_id[gene_inter_index],
                      tss_site,
                      sep = "__")
    } else {
      tss_site <- ifelse(strand == "+", tss_p2, tss_p1)
      mir_end <- ifelse(strand == "+", stem_loop_p2, stem_loop_p1)
      gene_tss_trans_index <- (tss_site - gene_df$gene_start) *
        (mir_end - gene_df$gene_start) < 0

      if (any(gene_tss_trans_index)) {
        distance <- abs(tss_site - gene_df$gene_start)
        gene_inter_index <- which.min(distance)
        result <- paste("inter__overlap_inter_TSS2",
                        gene_df$gene_id[gene_inter_index],
                        tss_site,
                        sep = "__")
      } else {
        gene_end <- ifelse(gene_df$strand == "+", gene_df$gene_p2,
                           gene_df$gene_p1)
        gene_end_trans_index <- (tss_site - gene_end) * (mir_end - gene_end) < 0
        if (any(gene_end_trans_index)) {
          select_gene <- gene_df[gene_end_trans_index, ]
          distance <- abs(tss_site - select_gene$gene_start)
          gene_inter_index <- which.min(distance)
          result <- paste("inter__overlap_inter_TSS",
                          select_gene$gene_id[gene_inter_index],
                          tss_site,
                          sep = "__")
        } else {
          distance <- abs(tss_site - gene_df$gene_start)
          gene_inter_index <- which.min(distance)
          result <- paste("inter__inter_TSS",
                          gene_df$gene_id[gene_inter_index],
                          tss_site,
                          sep = "__")
        }
      }
    }
  }
  result
}


tss_filter <- function(mir_name, chrom, stem_loop_p1, stem_loop_p2,
                       strand, tss_p1, tss_p2, expressed_gene = "all",
                       allmirgene_byforce = TRUE) {

  candidate_tss <- tibble::tibble(mir_name = mir_name, chrom = chrom,
                              stem_loop_p1 = stem_loop_p1,
                              stem_loop_p2 = stem_loop_p2,
                              strand = strand, tss_p1 = tss_p1,
                              tss_p2 = tss_p2)

  if (length(expressed_gene) > 1 || expressed_gene != "all") {
    gene_loci <- gene_loci %>%
      filter(gene_id %in% expressed_gene)
  }

  tss_info <- candidate_tss %>%
    select(stem_loop_p1, stem_loop_p2, chrom, strand, tss_p1, tss_p2) %>%
    pmap(tss_single_judge2, gene_loci) %>%
    unlist()

  all_mir_tss <- candidate_tss %>%
    bind_cols(tss_info = tss_info)

  s_tss <- all_mir_tss %>%
    select(-tss_p1, -tss_p2) %>%
    extand_df("tss_info", ":") %>%
    separate(new_info,
             into = c("mir_context", "tss_type", "gene", "predicted_tss"),
             sep = "__") %>%
    mutate(predicted_tss = as.double(predicted_tss))

  s_tss_success <- s_tss %>%
    filter(tss_type %in%
             c("host_TSS", "intra_TSS", "overlap_inter_TSS", "inter_TSS"))
  a1 <- s_tss_success$stem_loop_p1 - s_tss_success$predicted_tss
  a2 <- s_tss_success$predicted_tss - s_tss_success$stem_loop_p2
  pri_tss_distance <- ifelse(s_tss_success$strand == "+", a1, a2)

  result <- bind_cols(s_tss_success, pri_tss_distance = pri_tss_distance) %>%
    group_by(mir_name) %>%
    filter(pri_tss_distance == min(pri_tss_distance)) %>%
    ungroup()

  fail_mir <- setdiff(unique(candidate_tss$mir_name), unique(result$mir_name))

  if (length(fail_mir) == 0) {
    fail <- "All miRNAs have tss"
  } else {
    fail <- fail_mir
  }

  if (allmirgene_byforce == TRUE && length(fail_mir) != 0) {
    a_tmp <- s_tss %>%
      filter(mir_name %in% fail)
    aa1 <- a_tmp$stem_loop_p1 - a_tmp$predicted_tss
    aa2 <- a_tmp$predicted_tss - a_tmp$stem_loop_p2
    pri_tss_distance2 <- ifelse(a_tmp$strand == "+", aa1, aa2)
    fail_res <- bind_cols(a_tmp, pri_tss_distance = pri_tss_distance2) %>%
      group_by(mir_name) %>%
      filter(pri_tss_distance == min(pri_tss_distance)) %>%
      ungroup()

    fail_res$tss_type <- str_remove(fail_res$tss_type, "2")

    result <- bind_rows(result, fail_res)
  }

  result <- result[!duplicated(result), ]

  list(tss_df = result, fail_genefilter = fail)
}