Skip to contents

Extract modification-centered signal fragments from a poly(A) tail

Source: R/ninetails_core_functions.R
split_tail_centered.Rd

Finds areas in the poly(A) tail signal containing potential non-adenosine residues and extracts 100-point signal fragments where the potential modification is always at the center of a given extracted fragment.

Usage

split_tail_centered(readname, tail_feature_list)

Arguments

readname

Character string. Name of the given read (UUID) within the analyzed dataset.

tail_feature_list

List object produced by create_tail_feature_list. Must contain per-read entries with tail_signal, tail_moves, and tail_pseudomoves.

Value

A nested list where each element represents one candidate modification region. Each element is itself a list with:

chunk_sequence

Numeric vector of raw signal values (length 100)

chunk_start_pos

Integer. Start index of the chunk in the original signal

chunk_end_pos

Integer. End index of the chunk in the original signal

chunk_moves

Numeric vector. Basecaller moves for the chunk region

Element names follow the pattern <readname>_<chunk_index>. Positions are indexed from the 3' end.

Details

Candidate modification regions are identified based on two assumptions: the presence of significant raw signal distortion (recorded as a pseudomove by the thresholding algorithm) and the transition of state (move == 1) recorded by Guppy. If only move == 0 values are present within a given signal chunk, then that chunk is dropped from the analysis (the distortion is most likely caused by a sequencing artifact, not a non-A residue itself).

If the data indicating the presence of modifications are near the signal ends (3' or 5'), missing upstream or downstream data are imputed based on the most frequent values in the entire signal.

The extraction procedure is as follows:

  1. Run-length encoding (RLE) of the pseudomove vector

  2. Filtering runs of pseudomoves with length >= 5

  3. Centering a 100-point window on the midpoint of each qualifying run

  4. Imputing NAs at boundaries with draws from the 5 most frequent signal values

  5. Removing chunks where basecaller moves are all zero (likely artifacts)

See also

create_tail_feature_list for preparing the input, create_tail_chunk_list for batch segmentation, split_tail_centered_dorado for the Dorado-specific version

Examples

if (FALSE) { # \dontrun{

ninetails::split_tail_centered(
  readname = "1234-anexample-r3adn4m3",
  tail_feature_list = tail_feature_list
)

} # }