Skip to contents

Extracts tail features of single RNA read from respective basecalled multi-fast5 file.

Source: R/ninetails_training_dataset_production_functions.R
extract_tail_data_trainingset.Rd

This is the training-set variant of the single-read extraction function. It reads raw signal, basecaller moves and channel metadata from a multi-Fast5 file, isolates the poly(A) tail region defined by nanopolish, winsorizes the signal, downsamples by linear interpolation (to 20% of original length), and computes pseudomoves via filter_signal_by_threshold_trainingset.

Usage

extract_tail_data_trainingset(
  readname,
  polya_summary,
  workspace,
  basecall_group
)

Arguments

readname

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

polya_summary

Data frame. The table containing data extracted from nanopolish and the sequencing summary (produced by extract_polya_data).

workspace

Character string. Full path of the directory containing basecalled multi-Fast5 files.

basecall_group

Character string ["Basecall_1D_000"]. Name of the level in the Fast5 file hierarchy from which the data should be extracted.

Value

A named list with four elements:

fast5_filename

Character. Name of the source Fast5 file.

tail_signal

Numeric vector. Winsorized and downsampled signal corresponding to the poly(A) tail region.

tail_moves

Numeric vector. Downsampled basecaller moves for the tail region.

tail_pseudomoves

Numeric vector. Pseudomove values (-1, 0, 1) indicating potential non-A deviations in the tail signal.

Always assign this returned list to a variable; printing the full list to the console may crash the R session.

Details

The function differs from its production counterpart (extract_tail_data) in that:

  • The signal is downsampled (interpolated) to 20% of its original length to speed up downstream training-set preparation.

  • Pseudomoves are computed using the training-set-specific threshold filter (filter_signal_by_threshold_trainingset).

  • The returned list omits production-only fields and instead provides elements suitable for chunk splitting and GAF creation.

See also

create_tail_feature_list_trainingset for the parallel wrapper that calls this function, extract_tail_data for the production counterpart, winsorize_signal for the signal clipping step, filter_signal_by_threshold_trainingset for pseudomove detection.

Examples

if (FALSE) { # \dontrun{

extract_tail_data_trainingset(
  readname = 'abc123de-fg45-6789-0987-6543hijk2109',
  polya_summary = polya_summary_table,
  workspace = '/path/to/folder/containing/multifast5s',
  basecall_group = 'Basecall_1D_000')

} # }