Splits signal to overlapping fragments of equal length.

Divides a single read's poly(A) tail signal into fixed-length, overlapping segments for data augmentation during training-set preparation. If the signal is not evenly divisible by the segment length, trailing NA values are imputed with values randomly sampled from the 10 most frequent signal values.

Usage

split_with_overlaps(readname, tail_feature_list, segment, overlap)

Arguments

readname

Character string. Name (UUID) of the given ONT signal.

tail_feature_list

List object produced by create_tail_feature_list_A.

segment

Numeric [1]. Length (in data points) of each chunk to be created.

overlap

Numeric [1]. Length (in data points) of the overlap between consecutive chunks.

Value

A list of numeric vectors, each of length segment, representing the overlapping signal fragments. Trailing NA values are replaced by imputed values.

Details

This function is used exclusively in the A-only training pipeline (create_tail_chunk_list_A) to produce overlapping windows from reads that do not contain modification signals. The overlap acts as data augmentation, multiplying the number of training examples per read.

See also

create_tail_chunk_list_A which calls this function, create_tail_feature_list_A for the preceding step, split_tail_centered_trainingset for the modification-centered splitting alternative.

Examples

if (FALSE) { # \dontrun{

split_with_overlaps(
  readname = '12fdcb3-ewfd543-34552-1234ddta345',
  tail_feature_list = tail_feature_list,
  segment = 300,
  overlap = 100)

} # }