Skip to content

CHANGELOG.md

Latest commit

 

History

History
188 lines (150 loc) · 7.35 KB

CHANGELOG.md

File metadata and controls

188 lines (150 loc) · 7.35 KB
Raw

Changelog

All notable changes to the aa-tRNA-seq pipeline are documented in this file.

[Unreleased]

Fixed

  • Pipeline summary outputs (bcerror, odds_ratios) now report positions in tRNA-only coordinates (1-indexed) instead of full-reference coordinates that included adapter sequences. This fixes incorrect nucleotide positions in downstream tools like clover's plot_tRNA_structure().

Added

  • trim_reference rule produces a tRNA-only FASTA (trna_only.fa) by stripping 5'/3' adapter sequences from the adapted reference. This FASTA is used by clover for MODOMICS annotation and structure visualization.
  • build_trna_reference.py --mode trim for generating adapter-stripped tRNA-only FASTA files.
  • get_bcerror_freqs.py and compute_odds_ratios.py accept --offset-5p and --offset-3p to filter adapter positions and convert to tRNA-only coordinates.

Changed

  • Updated dorado from 1.3.1 to 1.4.0, basecalling model from v5.1.0 to v5.3.0, and modkit from 0.6.0 to >=0.6.1.
  • Shell scripts (setup-env.sh, setup-tools.sh) now read dorado_version and dorado_model from config/config-base.yml instead of hardcoding defaults.
  • Updated stale documentation references for dorado and modkit paths/versions.
  • Pipeline now warns on startup if installed dorado version does not match dorado_version in config.

[v0.1.1] - 2026-02-11

Fixed

  • bwa_align OOM at 48 GB: decoupled dorado tags from alignment by stripping tags from FASTQ, dropping bwa mem -C, and injecting tags from the unaligned BAM afterward via new inject_ubam_tags rule

Added

  • Quarto QC report with per-sample tabs (#81)
  • Per-tRNA pairwise modification odds ratios (#85)
  • Reference sequence similarity QC (#84)
  • Squiggy session JSON export for Positron IDE
  • Utility to collapse redundant GtRNAdb FASTA sequences
  • Multiple 3' adapter support for PT tag detection
  • Skip mode for reference validation
  • Pre-download dorado mod base models rule (avoids race conditions)
  • nvitop GPU monitoring dependency

Changed

  • classify_charging switched from GPU to CPU with parallel workers (8 threads)
  • WarpDemuX workflow simplified: eliminated merge_pods_for_demux, passes raw POD5 dirs directly
  • bwa_align filtering changed from -F 4 to -F 20 (also excludes reverse-strand reads)
  • Removed redundant awk position filter from bwa_align
  • Removed protected() directive from rebasecall output

Fixed

  • Race condition when parallel GPU jobs download dorado modification models simultaneously
  • Reverse-strand reads not filtered at alignment step
  • Redundant awk position filter in bwa_align superseded by adapter-based filtering
  • Graceful fallback for get_pipeline_commit when git unavailable
  • Various snakefmt formatting and test corrections

[v0.1.0] - 2025-01-16

Added

  • Run manifest generation for reproducibility tracking
  • Native SLURM cluster support with GPU configuration
  • Pytest unit tests and expanded CI coverage (#80)
  • Adapter position tagging with parasail alignment - adds PT tags (#79)
  • MkDocs documentation website (#77)
  • Publication citation (White et al. 2025 Nat Commun)
  • Reference validation and building step to ensure tRNA sequences have proper CCA endings and adapter structure required for charging classification
  • Optional WarpDemuX barcode demultiplexing support for pooled/multiplexed sequencing runs (#74)
  • Optimized modkit thresholds from ModkitOpt
  • Pixi package manager support as primary environment manager
  • Mermaid diagram for workflow visualization

Changed

  • Standardized output directory structure to nested sample paths
  • Updated dorado version from 0.9.1 to 1.3.1
  • Migrated GitHub Actions CI from conda to pixi (#75)
  • Separated tool installation from environment activation (pixi run setup)
  • Updated README to use pixi instead of conda (#72)
  • Normalized file permissions across repository

Fixed

  • Major fix: Improved 5' adapter detection from 0.04% to 82% (#82)
  • Resolved GLIBCXX version errors with configurable CUDA support
  • Shell glob pattern errors in tests

2025-11-07

Added

  • Claude Code session start hook for automated development setup (#69)
  • Comprehensive CI/CD build and test checks (#68)
  • New project initialization structure (#65)

Changed

  • Prefer pandas over polars for stability on some cluster nodes
  • Reduced Remora logging level

2025-06-22

Added

  • LSF-specific cluster configuration (#63)

Changed

  • Renamed test directory to recommended .tests/ location
  • Reduced and made optional dorado verbosity
  • Downgraded numpy to fix remora stats compatibility

2025-03-16

Added

  • Modkit integration for RNA modification analysis (#59)
    • modkit_pileup rule for modification pileups
    • modkit_summary rule for modification summaries
    • modkit_extract and modkit_extract_full rules for detailed modification data
  • Automatic dorado and model download/installation (#56)
  • Modified base calling support (pseU, m5C, inosine_m6A)
  • Full modkit outputs with optimized memory allocation

Changed

  • Eliminated support for FAST5 files - pipeline now POD5-only (#43)
  • Reorganized output directory structure
  • Renamed charging tags during transfer (ML→CL, MM→CM)
  • Updated model download strategy
  • Increased memory allocation for modkit rules

Fixed

  • Restored -v option in dorado for proper verbosity control

2025-01-08

Added

  • Rule for calculating CPM of charged/uncharged tRNAs (#28)
  • Remora CCA classifier for charging state classification (#18)
  • GPU pipeline support for cca_classify rule (#23)
  • Charging probability extraction and analysis (#46)

Changed

  • Implemented ML threshold (≥200 = charged, <200 = uncharged)
  • Compressed output files for storage efficiency
  • Various tweaks to file handling (#27)

Fixed

  • Actually use the threshold value in classification

2024-08-13

Added

  • Alignment filtering capabilities with configurable parameters (#13)
  • Optional Remora signal metrics extraction
  • Kmer models included in pipeline
  • Logging and optional failed BAM outputs
  • New BAM tag indicating why reads are filtered
  • Support for processing reads from both pass and fail directories

Changed

  • Cleanup filtering approach for full-length tRNA reads
  • Updated test data
  • Ignore supplementary and secondary alignments
  • Snakemake v8 compatibility (#10)

Fixed

  • Insertion double-counting bug (#15)
  • Dropped redundant summary align stats (#14)
  • Added 'pod5' to list of possible pod5 directories

2024-05-19

Added

  • Support for merging multiple sequencing runs per sample
  • Support for unmapped BAM as input (#5)
  • Pipeline commit and config recording for reproducibility (#9)
  • Bedgraph output generation
  • Alignment statistics calculations
  • Base calling error frequency calculations

Changed

  • Use v5.0.0 dorado models with modification calling
  • Expose dorado and bwa command-line options
  • Reworked alignment stats output (#8)
  • Set rebasecalled outputs as read-only

Fixed

  • Keep additional BAM flags (e.g., pi) during processing (#1)
  • Use -T → -C options to preserve all BAM tags from dorado

2024-02-07

Added

  • Initial pipeline release
  • Core workflow: POD5 merge → rebasecall → align → filter
  • BWA MEM alignment to tRNA + adapter reference
  • Post-alignment filtering for full-length tRNAs
  • Basic summary statistics generation
  • Snakemake workflow with modular rule structure
  • Conda environment specification
  • Sample configuration via TSV files