Skip to contents

Extracts decoration-centered fragments of poly(A) tail signal along with positional coordinates.

Source: R/ninetails_training_dataset_production_functions.R
split_tail_centered_trainingset.Rd

Splits a single read's poly(A) tail signal into fixed-length (100 data-point) chunks, each centered on a potential non-A decoration detected from the pseudomove vector. Chunks whose start position falls before index 1 are left-padded with values randomly sampled from the three most frequent signal values.

Usage

split_tail_centered_trainingset(readname, tail_feature_list)

Arguments

readname

Character string. Name (UUID) of the given read.

tail_feature_list

List object produced by create_tail_feature_list_trainingset.

Value

A named nested list where each element corresponds to one chunk and contains:

chunk_sequence

Numeric vector (length 100). The signal fragment centered on the potential modification.

chunk_start_pos

Integer. Start index of the chunk within the full tail signal (may be negative for left-padded chunks).

chunk_end_pos

Integer. End index of the chunk within the full tail signal.

pseudomoves

Numeric vector (length 100). Recomputed pseudomoves for the chunk. Coordinates are from the 3' end.

Details

This training-set variant includes an additional pseudomoves element in each chunk sublist, making the output suitable for supervised training/validation data preparation. Pseudomoves for each extracted chunk are recomputed by calling filter_signal_by_threshold on the chunk sequence.

The centering procedure:

  1. Runs RLE on the pseudomove vector.

  2. Selects runs of length >= 4 with non-zero values (empirical modification threshold).

  3. Centers a 100-element window on the midpoint of each selected run.

Chunk names follow the convention <readname>_<index>, where index is the sequential position of the modification within the read (numbered from the 3' end).

See also

create_tail_chunk_list_trainingset for the parallel wrapper that calls this function, filter_signal_by_threshold for pseudomove recomputation on individual chunks, split_tail_centered for the production counterpart.

Examples

if (FALSE) { # \dontrun{

split_tail_centered_trainingset(
  readname = "1234-anexample-r3adn4m3",
  tail_feature_list = tail_feature_list)

} # }