format_vcf_fields_longer.R

#!/usr/bin/env Rscript

# Write stdout and stderr to log file
log <- file(snakemake@log[[1]], open = "wt")
sink(log, type = "message")
sink(log, type = "output")

library(readr)
library(dplyr)
library(tidyr)
library(stringr)
library(purrr)
library(rlang)
library(glue)
library(logger)

empty.to.na <- function(x) {
  x[x == ""] <- NA
  x
}

# Replace "" values with NA in R filter list
# Snakemake passes filters like: list(ERRORS = c(""))
log_info("Preprocessing data")
filter.include <- lapply(snakemake@params$filter_include, empty.to.na)
filter.exclude <- lapply(snakemake@params$filter_exclude, empty.to.na)

# Process input table
log_info("Applying filters and writing results")
df <- read_tsv(
  snakemake@input$tsv,
  col_types = cols(
    POS = col_integer(),
    .default = col_character()
  )
) %>%

  # Separate <sep>-delimited "...[*]..." columns (e.g. ANN[*].EFFECT)
  separate_rows(
    contains("[*]"),
    sep = snakemake@params$sep,
    convert = FALSE
  ) %>%

  # Rename "...[*]..." columns using the provided lookup via Snakemake config
  rename(all_of(unlist(snakemake@params$colnames_mapping)))

if (nrow(df) == 0) {
  log_info("Writing empty file")
  write_tsv(df, snakemake@output$tsv)
} else {
  log_info("Processing variants")
  df %>%
    # Ensure missing values are properly encoded
    mutate(across(where(is.character), ~ na_if(.x, "NA"))) %>%

    # Separate &-delimited error column (more than one error/warning/info message per row is possible)
    mutate(split_errors = strsplit(ERRORS, "&")) %>%
    # Keep rows with none of the excluded ERRORS terms, if any
    filter(map_lgl(split_errors, ~ !any(. %in% filter.exclude[["ERRORS"]]))) %>%
    select(-split_errors) %>%

    # Apply filters
    filter(
      # Keep variants that include required values in each field
      !!!map2(
        names(filter.include),
        filter.include,
        ~ expr(.data[[!!.x]] %in% !!.y)
      ),
      # Keep variants that exclude required values in each field
      !!!map2(
        names(filter.exclude),
        filter.exclude,
        ~ expr(!(.data[[!!.x]] %in% !!.y))
      )
    ) %>%

    # Keep unique rows
    distinct() %>%

    mutate(
      # Assign variant name using the pattern defined via Snakemake config
      VARIANT_NAME = str_glue(snakemake@params$variant_name_pattern),
      # Assign sample name
      SAMPLE = snakemake@params$sample
    ) %>%

    # Write output file
    write_tsv(snakemake@output$tsv)

  log_info("Done")
}