Skip to contents

Marks uncertain positions of non-A residues in ninetails output data.

Source: R/ninetails_data_postprocessing_functions.R
correct_residue_data.Rd

Annotates each non-A residue position with a quality flag (qc_pos) indicating whether the position is likely genuine ("Y") or a potential nanopolish segmentation artefact ("N"). The assessment is based on quantile thresholds of poly(A) tail length, modal non-A position per transcript, and optional species-specific whitelists of transcripts with A-rich 3' UTRs.

Usage

correct_residue_data(
  class_data,
  residue_data,
  grouping_factor = NULL,
  transcript_column,
  ref = NULL
)

Arguments

class_data

Data frame or tibble containing read_classes predictions from the ninetails pipeline.

residue_data

Data frame or tibble containing non-A residue predictions from the ninetails pipeline.

grouping_factor

Character string or NULL (default). A grouping variable (e.g. "sample_name", "group").

transcript_column

Character string. Name of the column containing transcript identifiers (e.g. "ensembl_transcript_id_short").

ref

Character string, character vector, or NULL (default). Whitelist of transcripts with hybrid tails. Built-in options:

"athaliana"

Arabidopsis thaliana

"hsapiens"

Homo sapiens

"mmusculus"

Mus musculus

"scerevisiae"

Saccharomyces cerevisiae

"celegans"

Caenorhabditis elegans

"tbrucei"

Trypanosoma brucei

A custom character vector of transcript IDs may also be provided. Must be consistent with the content of transcript_column.

Value

A tibble based on residue_data with the following additional columns:

mode_pos

Integer. Most frequent non-A position reported for the transcript.

mode_len

Integer. Most frequent tail length reported for the transcript.

seg_err_quart

Numeric. 0.05 quantile of tail length for the transcript.

qc_pos

Character. Quality flag: "Y" for likely genuine, "N" for ambiguous.

pos_err_quart

Numeric. 0.05 quantile of non-A position for the transcript and prediction type.

count_nonA

Integer. Number of non-A-containing reads for the transcript.

count

Integer. Total number of reads for the transcript.

Details

Nanopolish segmentation can misidentify nucleotides from A-rich 3' UTR regions as part of the poly(A) tail. For such transcripts, a peak of non-A positions accumulates near the transcript body. This function flags those positions as ambiguous using a combination of:

  • The modal position of non-A residues per transcript.

  • The 0.05 quantile of poly(A) tail lengths per transcript (segmentation error boundary).

  • The 0.05 quantile of non-A positions per transcript and prediction type.

  • Species-specific whitelists of transcripts with hybrid tails (3' UTRs with > 80% A in the last 20 positions).

See also

correct_class_data for the companion function that reclassifies reads based on this output, reclassify_ninetails_data for the high-level wrapper, check_tails_guppy and create_outputs for the pipeline that produces the input data.

Examples

if (FALSE) { # \dontrun{

residue_data_edited <- ninetails::correct_residue_data(
  class_data = results[[1]],
  residue_data = results[[2]],
  transcript_column = "contig")

} # }