Skip to content

Changelog

2.5.0 — March 2026

🐳 Docker images

Assembly Image
hg38 ghcr.io/friendsofstrandseq/mosaicatcher-pipeline:hg38-2.5.0
hg19 ghcr.io/friendsofstrandseq/mosaicatcher-pipeline:hg19-2.5.0
T2T ghcr.io/friendsofstrandseq/mosaicatcher-pipeline:T2T-2.5.0
mm10 ghcr.io/friendsofstrandseq/mosaicatcher-pipeline:mm10-2.5.0
mm39 ghcr.io/friendsofstrandseq/mosaicatcher-pipeline:mm39-2.5.0

📋 Summary

This release focuses on pipeline robustness: corrupted BAMs are now caught immediately after alignment and merging, StrandPhaseR no longer crashes on chromosomes with insufficient SNPs, and a Snakemake v9 checkpoint deadlock in the StrandPhaseR step has been resolved. SLURM resource requests across memory, CPU threads, and runtimes have been right-sized based on empirical data from ~9,700 jobs.

✨ Features

  • AVITI watcher — new Snakemake v9 Python API script (workflow/scripts/toolbox/watcher/) to monitor AVITI sequencing runs and trigger pipeline execution automatically
  • Shared cache setup — setup_shared_caches.sh creates group-writable Apptainer image and conda environment directories on EMBL HPC with correct ACLs

🐛 Fixes

  • Assembly binning — bin BED files and chromosome lists are now resolved per reference assembly; previously all assemblies fell back to the primary reference, causing incorrect binning for hg19, T2T, mm10, mm39
  • StrandPhaseR checkpoint deadlock — resolved a Snakemake v9 deadlock where the jobstep checkpoint caused StrandPhaseR jobs to hang indefinitely
  • StrandPhaseR chromosome guard — chromosomes with fewer SNPs than StrandPhaseR's minimum are now skipped with a logged warning instead of crashing with a cryptic R error
  • Corrupted BAM detection (alignment) — samtools quickcheck added after alignment group; truncated BAMs from node failures are caught before downstream steps
  • Corrupted BAM detection (merge) — samtools quickcheck added after mergeBams and mergeSortBams; prevents silent propagation of corrupt merged files on resume
  • SLURM profile runtime comment — default-resources: runtime=600 was documented as "minutes" but is in seconds (= 10 min); comment corrected

⚙️ Misc

  • Right-sized SLURM resources (based on empirical data from 9,700+ jobs across 22 runs):
  • create_haplotag_table memory: 8 GB → 12 GB base (p95 observed: 9.7 GB, max: 17.4 GB)
  • mosaiClassifier_calc_probs memory: 8 GB → 2.5 GB base (max observed: 2.8 GB)
  • run_strandphaser_per_chrom memory: 8 GB → 4 GB base; runtime: 600 → 120 min
  • call_SNVs_bcftools_chrom memory: 8 GB → 1 GB base; runtime: 180 → 30 min
  • mosaiClassifier_calc_probs runtime: 180 → 60 min
  • ashleys_generate_features threads: 64 → 12; runtime: 3600 → 60 min
  • ploidy_estimation threads: 48 → 24

2.4.0 — February 2026

🐳 Docker images

Assembly-specific images introduced in this release. Each image embeds the matching BSgenome R package:

Assembly Image
hg38 ghcr.io/friendsofstrandseq/mosaicatcher-pipeline:hg38-2.4.0
hg19 ghcr.io/friendsofstrandseq/mosaicatcher-pipeline:hg19-2.4.0
T2T ghcr.io/friendsofstrandseq/mosaicatcher-pipeline:T2T-2.4.0
mm10 ghcr.io/friendsofstrandseq/mosaicatcher-pipeline:mm10-2.4.0
mm39 ghcr.io/friendsofstrandseq/mosaicatcher-pipeline:mm39-2.4.0

📋 Summary

2.4.0 is a major infrastructure release: the pipeline migrates to Snakemake v9 and Pixi for reproducible environment management, the ashleys-qc-pipeline is fully integrated (no longer a submodule), and all hardcoded genome references are replaced by a centralized genome registry. New assemblies (mm39, CanFam) and a new EMBL HPC Apptainer profile with shared container caches are included.

✨ Features

  • Snakemake v9 + Pixi — unified package management via pixi.toml; pixi run snakemake replaces all conda-activate workflows
  • Assembly-specific containers — one GHCR image per reference genome embedding the matching BSgenome R package; automated multi-assembly container builds in CI
  • Centralized genome registry — all hardcoded genome paths replaced by a single references_data config block; adding a new assembly now requires a single config entry
  • ashleys-qc-pipeline integration — preprocessing pipeline no longer a git submodule; lives directly in workflow/rules/ashleys/
  • mm39 full support — normalization files, blacklist regions, BSgenome package for mouse GRCm39
  • CanFam framework — reference infrastructure in place for canfam3/canfam4 (normalization files pending)
  • Pre-built iGenomes index download — download_prebuilt_indexes parameter: skip local BWA index building by downloading from AWS iGenomes
  • Ploidy estimation module — ploidy parameter: optional detection of haploid chromosomes/segments to guide StrandPhaseR
  • Centralized version management — pixi run bump-patch|bump-minor|bump-major|bump-beta|bump-release; automated changelog generation for releases
  • EMBL HPC Apptainer profile — new profile at workflow/snakemake_profiles/mosaicatcher-pipeline/v9/HPC/slurm_EMBL_apptainer/ with shared container caches, SLURM job grouping, and resource efficiency reporting
  • Configurable reference_base_dir — multi-user reference sharing on HPC without per-user downloads
  • SLURM job grouping — horizontal (batch short jobs) and vertical (chain sequential steps) grouping to reduce scheduler overhead

🐛 Fixes

  • Empty DataFrames in clustering plots no longer raise errors
  • Assertion added for iGenomes base URL availability before attempting index download
  • Redundant index_input_bam rule removed
  • regenotype_SNVs handles empty site lists correctly
  • Conda environments cleaned: anaconda/defaults channels removed across all envs

⚙️ Misc

  • keep_ashleys_predictions parameter to control retention of ashleys ML prediction files
  • external_data_v8.smk renamed to external_data_v9.smk
  • snakemake_profiles submodule replaced with local versioned files
  • CI matrix expanded: mm39 indexed and tested; T2T with latency-wait for storage

2.3.0 — July 2024

🐳 Docker images

ghcr.io/friendsofstrandseq/mosaicatcher-pipeline:2.3.x

📋 Summary

2.3.0 adds BreakpointR and WhatsHap as optional execution targets, enabling users to stop the pipeline at intermediate steps for manual inspection. Snakemake v7 and v8 are both tested in CI, and several conda environment and path issues are resolved.

✨ Features

  • BreakpointR integration — breakpointr: True parameter activates BreakpointR for SV calling; breakpointr_only: True stops execution after BreakpointR
  • whatshap_only parameter — stop execution after WhatsHap phasing for manual review before SV classification
  • Paired-end support improvements — paired_end parameter handling stabilised
  • Data reorganisation script — workflow/scripts/toolbox/reorganise_data/ for reformatting data from external sequencing providers (DKFZ)

🐛 Fixes

  • IGV session script: reference genome path correctly resolved
  • scNOVA_DL conda environment fixed (note: only functional outside EMBL due to network restrictions)
  • Missing publishdir function in common.smk restored
  • Conda environments: removed anaconda/defaults channels, migrated to conda-forge/bioconda only
  • SLURM job time formatting corrected
  • Duplicate column in clustering plot fixed (previously silently ignored by some ggplot2 versions)
  • Missing wildcards in grouptrack rule resolved

⚙️ Misc

  • CI: Snakemake v7 and v8 parallel testing
  • .tests-mock submodule added for faster CI without LFS
  • StrandPhaseR updated to bioconda package (removed custom install)