Skip to contents

Changelog

Source: NEWS.md

ninetails 1.0.84 (development)

cDNA pipeline: independent validation of orientation calls

The cDNA pipeline classifies each read as polyA, polyT or unidentified by matching Dorado’s SSP and VNP primers against both ends of the basecalled sequence. That call is made entirely from sequence, so nothing in the pipeline previously contradicted it when it was wrong.

  • infer_cdna_layout() reads the orientation off the signal instead, by comparing the sizes of the pre-tail and post-tail regions and taking the larger side to be the transcript body. It is deliberately strict — a region below min_region_samples (default 100), or a larger/smaller ratio below strictness_ratio (default 1.5), returns "ambiguous" rather than committing to an orientation.
  • cdna_layout_agreement_marker() pairs the sequence-based call with the layout-based one and collapses the result to a single character, so disagreements are visible at a glance in a table of thousands of reads.
  • launch_cdna_signal_browser() puts the two side by side. It takes a Dorado summary and the output of detect_orientation_multiple(), joins them on read_id, and reports how many reads carried no matching tail_type rather than silently dropping them.

The two methods are independent by construction — one reads primers, the other reads region geometry — so agreement is evidence and disagreement is a read worth looking at.

Dorado 1.4.0 compatibility

Dorado 1.4.0 renamed the POD5 column in its summary output from filename to input_filename. preprocess_inputs() and launch_cdna_signal_browser() now accept either spelling, normalising to filename internally.

Bug fixes

  • Read IDs containing underscores are no longer truncated. Chunk names have the form <read_id>_<chunk_index>, and the read ID was recovered by splitting on _, which cut UUID-style identifiers short and orphaned their predictions. Only the trailing _<digits> is stripped now.
  • detect_orientation_multiple() no longer overwrites its input files. The classified table is written to a sibling <name>_classified.tsv, leaving the original sequence files untouched. Overwriting them risked corrupting user input on a partial failure, and made re-runs non-idempotent — the second pass would re-classify an already-tagged file.
  • An empty signal vector no longer aborts feature extraction. filter_signal_by_threshold() returns no pseudomoves for a zero-length signal rather than iterating its z-score loop backwards.
  • split_tail_centered_dorado() pads chunks that would start before the beginning of the signal with random draws from the five most frequent observed values, rather than a fixed constant, keeping chunk length at 100 without introducing a flat artificial segment for the CNN to learn from.

Documentation

  • The package now has a changelog, rendered on the pkgdown site.
  • The pkgdown reference index groups all exported functions by pipeline and role, with the Guppy pipeline marked legacy throughout.
  • Vignettes covering detection, postprocessing, plotting, signal inspection, the Shiny dashboard and tailfindr compatibility.

ninetails 1.0.78

Released 2026-05-07.

Dorado-compatible DRS processing with a Shiny application for signal browsing.

  • launch_signal_browser() for Dorado DRS (POD5) data and launch_signal_browser_guppy() for Guppy legacy (fast5) data, with non-A modification overlays on the raw signal trace.
  • The cDNA pipeline is present but under construction and should not be used for analysis.

ninetails 1.0.4

Released 2025-09-28.

  • Full POD5 and Dorado support. check_tails_dorado_DRS() processes direct RNA sequencing data basecalled with Dorado ≥ 1.0.0, reading poly(A) coordinates from the Dorado summary rather than Nanopolish.
  • The fast5/Guppy pipeline is preserved as legacy mode via check_tails_guppy(), and is no longer actively developed.
  • Additional pipelines announced as forthcoming.

ninetails 1.0.3

Released 2024-12-05.

  • Fixed the export of the tailfindr compatibility function.

ninetails 1.0.2

Released 2024-04-02.

The v.1.0.2_manuscript tag (2024-08-13) marks the version accompanying the manuscript.

ninetails 1.0.0

Released 2023-11-23.

  • First stable release. Features added, code optimised, typos fixed.

ninetails 0.9.0

Released 2023-10-03.

  • Performance fixes.

ninetails 0.7.0

Released 2023-01-18.

  • Performance fixes.
  • Fixed potential issues arising from input file incompatibility.
  • Clarified comments and descriptions throughout.

ninetails 0.4.1

Released 2022-10-13.

  • Added statistical functions, data visualisation, and postprocessing features.
  • Fixed a bug that diverted stdout.

ninetails 0.4.0

Tagged 2022; no accompanying release notes.

ninetails 0.3.1

Released 2022-09-02.

  • Windows compatibility.

ninetails 0.3.0

Released 2022-08-23.

  • Fixed R dependency issues.
  • Fixed the Nanopolish input argument.

ninetails 0.2.0

Released 2022-08-23.

  • Prerelease with major updates.