📝 V1.32.0 Release Notes
Summary
🐛 - bug fix
📘 - documentation improvement.
🆕 - new feature.
🎉 - quality improvement.
EfficientDV
Functional changes
- 📘 how-to fixes (reference files + docker tags)
- 🆕 Added support for the hg38_no_alt genome (GCA_000001405.15_GRCh38_no_alt_analysis_set)
- 🎉 Adds support for the new version of hg38 that is based on the GIAB best practices hg38 version (GIABv3), but also contains the same decoy sequences as the Broad's version of hg38 (b38). This version of the genome is called
hg38_nistv3_wth_decoy. Using this genome in the variant calling improves the F1 for SNP calling for the GIAB samples by approximately 0.1-0.2% dependent on the sample. - 🆕 Replaced two-pass ensemble approach (CallVariantNoEnsemble + UGSplitBoundaryCalls + CallVariantsBoundary) with a single unified CallVariants task that handles selective ensemble internally
New Parameters
- active_areas_min_base_quality — Minimum base quality for active areas detection
- prioritize_high_quality_reads — When min_mapq=0, fills pileup image with high-mapq reads first before adding mapq=0 reads (default: true)
- trim_soft_clips — Trims soft-clipped bases from pileup images (default: true)
- v_gpu_tile_size — Virtual GPU tile size for call_variants
Parameter Changes (breaking)
- pileup_min_mapping_quality and candidate_min_mapping_quality merged into a single min_mapping_quality parameter controlling both pileup image inclusion and candidate generation
- call_variants_gpu_type (required) → call_variants_gpu_type_override (optional, defaults to nvidia-t4-a10g-l4)
Removed Outputs
call_variants_output_tfrecords_final — removed
num_weak_candidates / num_weak_candidates_as_int — removed (no longer produced by the unified inference path)
Make-examples release notes: 🎉 📘
- UGCallVariants now writes ensemble and GPU settings into params.ini (vGPUTileSize, builderOptimizationLevel, ensembleSize, threshold from strong_call_threshold, criteria = max_prob_threshold, etc.).
- Adds workflow input v_gpu_tile_size (default 4; somatic presets often 2).
- Drops outputs num_weak_candidates and candidate counting expects log line num_candidates: (4.1.0).
- Allow changing the image width in make_example
- Htslib CRAM/BAM access (when --htslib specified)
- separate min-base-quality thresholds
- update mapping quality parameters
GiraffeAlignment
- 🆕 Expose KMC memory as parameter in input
MRDFeatureMap
Functional changes
- how-to docs fixes (reference files + docker tags) 📘
API changes
- Filter PoN variants by exact alt-allele in tumor informed MRD. 🎉
PyPGx
- 🎉 Added support for the hg38_no_alt genome (GCA_000001405.15_GRCh38_no_alt_analysis_set). This is the genome that we recommend to use with PyPGx pipeline.
- 📘 how-to docs fixes (reference files + docker tags)
STRGenotyper
- 📘 how-to docs fixes (reference files + docker tags)
SVPipeline
- 🐛 In converted VCF output of SV pipeline, the SVLEN "number" value was 1 -> changed to "." per VCF standard.
See Ultimagen/gridss#32 - 📘 Added support for the hg38_no_alt genome (GCA_000001405.15_GRCh38_no_alt_analysis_set)
- how-to docs fixes (reference files + docker tags)
SegDupAnalysis
- 🎉 Calling variants and CNV on IKBKG and HBA. The model was trained on 1000 genomes data. See JIRA for the details of training and evaluation.
- Added support for the hg38_no_alt genome (GCA_000001405.15_GRCh38_no_alt_analysis_set)
- 📘 how-to docs fixes (reference files + docker tags)
SingleReadSNV
Functional changes
- solve snvfind regression 🐛
SomaticSNVFind
Functional changes
- solve snvfind regression 🐛
TrimAlignSort
Functional changes
- 📘 how-to fixes (reference files + docker tags)
- 🆕 Added support for the hg38_no_alt genome (GCA_000001405.15_GRCh38_no_alt_analysis_set)
- 🆕 Adds support for NIST best practices hg38 version (GIABv3) with the same decoy sequences as the Broad's version of hg38 (b38). This version of the genome is called
hg38_nistv3_wth_decoy.