Skip to contents

Preprocess Dorado inputs for ninetails analysis (no BAM processing)

Source: R/ninetails_core_functions_dorado_DRS.R
preprocess_inputs.Rd

This function prepares Dorado summary files and extracts poly(A) signals from POD5 files for downstream analysis. It splits large summary files into manageable parts and extracts corresponding poly(A) signals.

Usage

preprocess_inputs(
  dorado_summary,
  pod5_dir,
  num_cores,
  qc,
  save_dir,
  prefix,
  part_size,
  cli_log
)

Arguments

dorado_summary

Character or data frame. Path to Dorado summary file or data frame.

pod5_dir

Character. Directory containing pod5 files.

num_cores

Integer. Number of CPU cores to use.

qc

Logical. Whether to perform quality control filtering. When TRUE, applies the following stringent filters to ensure high-quality data:

Alignment direction

Removes unmapped reads (alignment_direction == "*")

Mapping quality

Removes reads with zero mapping quality (alignment_mapq == 0)

Poly(A) coordinates

Removes reads with invalid start positions (poly_tail_start == 0)

Tail length

Removes reads with tails shorter than 10 nucleotides (poly_tail_length < 10)

These filters remove low-quality reads, misaligned sequences, and likely artifacts from pore clogging or basecalling errors. Set to FALSE only for preliminary analysis or when using pre-filtered data.

save_dir

Character. Directory where output files will be saved.

prefix

Character. Prefix to add to output file names (optional).

part_size

Integer. Number of reads to process in each chunk.

cli_log

Function for logging messages and progress.

Value

List containing paths to processed files:

summary_files

Paths to split summary files

polya_signal_files

Paths to extracted poly(A) signal files

Quality Control Details

The quality control filtering (when qc = TRUE) is implemented via filter_dorado_summary and performs the following operations:

  • Unmapped read removal: Reads with alignment_direction = "*" indicate failed alignment to the reference and are excluded.

  • Low mapping quality removal: Reads with alignment_mapq = 0 have ambiguous or poor-quality alignments and are excluded.

  • Invalid coordinate removal: Reads with poly_tail_start = 0 indicate failed poly(A) detection, often from pore clogging artifacts.

  • Short tail removal: Reads with poly_tail_length < 10 nucleotides lack sufficient poly(A) sequence for reliable modification detection.

These criteria ensure that only high-confidence reads with valid poly(A) tails and proper genomic alignment are processed, significantly improving downstream classification accuracy and reducing false positive rates.

Performance Note

Filtering typically removes 25-30% of raw reads from standard Dorado output, but this percentage may vary based on sequencing quality, pore condition, and alignment parameters. The function logs the exact number of reads removed for quality control tracking.

Examples

if (FALSE) { # \dontrun{
processed_files <- preprocess_inputs(
  dorado_summary = "path/to/summary.txt",
  pod5_dir = "path/to/pod5/",
  num_cores = 2,
  qc = TRUE,
  save_dir = "path/to/output/",
  prefix = "experiment1",
  part_size = 40000,
  cli_log = message
)
} # }