3.2.0 - Bebenhausen

Added

• #362 Optional germline_vcf samplesheet column: the patient's germline calls are folded into the neoepitope windows as context, so wild-type and mutant sequences are the patient's own rather than the reference. Germline records are never scanned for candidates (@axelwalter)
• #362 Replaced the epytope/BioMart variant path with an offline chain of bcftools, Ensembl VEP and pVACtools generate_protein_fasta. New params --ref_fasta, --vep_cache, --vep_species, --vep_genome, --vep_cache_version, --mutation_flanking_aas and --vep_download_cache, plus an optional tumor_sample samplesheet column. VCFs without a GT field (Strelka) are accepted, with the tumour genotype filled in by bcftools +setGT, following the approach proposed and tested in #373 (@axelwalter, @Ngarciar24)
• #362 Variant peptides are restricted to k-mers overlapping the mutation, cut per peptide length as in pvacseq run, and carry provenance (gene, transcript, consequence, HGVSp, genomic anchor, UniProt). Frameshifts now extend to the new stop codon instead of being truncated at the original protein length (@axelwalter)
• #362 --vep_cache also accepts a .tar.gz, which lets the test profile run the variant path on a 15 MB GRCh38 chr4+chr19 cache subset (@axelwalter)
• #362 test_grcm39 profile: the variant path on a 4 MB mouse GRCm39 chr15/18/19 fixture (@axelwalter)
• #362 Nearby somatic missense variants on the same transcript are also combined into one mutant sequence (assumed in cis unless the VCF carries FORMAT/HP phasing), so multi-variant peptides are generated alongside the single-variant ones (@axelwalter)
• #362 Kept --proteome_reference: variant peptides found in the given reference proteome are dropped before prediction (@axelwalter)
• #333 Added metro map to README
• #327 Added optional parameter use_ba_rank to prefer BA_Rank as rank metric in output of netmhc predictions (@jonasscheid)
• #341 Added <species>-all alleles (e.g. HLA-all) and allele chunking, download MHCflurry models once per run (fixes #340) (@jonasscheid)
• #338 Added MixMHCpred (Class I) and MixMHCIIpred (Class II) binding predictors, built on the fly with Wave and gated by --accept_mixmhcpred_license (@jonasscheid)
• #377 Added AGENTS.md from the nf-core pipeline template (@jonasscheid)
• #372 --wild_type predicts the wild-type counterparts of variant peptides (@jonasscheid)

Fixed

• #372 Fixed --wide_format_output dropping scores of peptides with missing metadata (@jonasscheid)
• #364 Fixed inverted --proteome_reference self-filtering that retained self-epitopes instead of removing them (fixes #363) (@jonasscheid)
• #368, #341 Fixed NetMHCpan and NetMHCIIpan crashing with a buffer overflow when run from a long working directory (@jonasscheid)
• #359 Fixed shuffled NetMHCpan/NetMHCIIpan allele labels by reading allele names from the xls header (fixes #358) (@jonasscheid)
• #352 Accept 3- and 4-field HLA typings by truncating to 2 fields via mhcgnomes (fixes #350) (@jonasscheid)
• Fixed NetMHCIIpan mouse class II allele conversion: H2-AA*b/AB*b now maps to H-2-IAb (and H2-EA*d/EB*d to H-2-IEd) instead of the invalid H-2-AAb-ABb (@jonasscheid)
• #349 Fixed MultiQC stats mismatch: use per-peptide binder counts, fix unsupported count formula, fix netmhcpan multi-allele column parsing (@jonasscheid)
• #331 Fixed Nextflow strict syntax lint errors (@jonasscheid)
• #348 Bumped Python container from 3.11 to 3.14 in SPLIT_PEPTIDES (@jonasscheid)

Dependencies

Dependency	Old version	New version
bcftools	1.21	1.23.1
multiqc	1.31	1.35
python	3.11.0	3.14.3
nf-core	3.4.1	4.1.0
NetMHCIIpan	4.3e	4.3
MixMHCpred		3.0
MixMHC2pred		2.0.2
ensembl-vep		116.1
pvactools		7.0.1
epytope	3.3.1
snpsift	4.3.1t

Parameters

Old parameter	New parameter
	--accept_mixmhcpred_license
--fasta_peptide_flanking_region_size	--mutation_flanking_aas
	--ref_fasta
	--use_ba_rank
	--vep_cache
	--vep_cache_version
	--vep_download_cache
	--vep_genome
	--vep_species
--fasta_output
--genome_reference
--hook_url
--netmhc_system
--split_by_variants
--split_by_variants_distance
--split_by_variants_size

Changed

• #336 Added NetMHCpan 4.2 mode config file instructions to the nf-core usage documentation. (@Kabooni).
• #346 Replace CAT_CAT with FIND_CONCATENATE to fix output filename collisions (@jonasscheid).
• Bump nf-core modules and subworkflows to latest (@jonasscheid).
• #361 Bump nf-schema plugin to 2.7.2 (@jonasscheid).
• #365 Merge nf-core template updates up to 4.0.3 (@jonasscheid).
• #371 Merge nf-core template updates up to 4.1.0 (@jonasscheid).
• #379 Check the NetMHCpan and NetMHCIIpan version from the tarball's data/version during pipeline initialisation instead of md5 checksums, so any sub-release of NetMHCpan 4.2 and NetMHCIIpan 4.3 is accepted (fixes #378) (@jonasscheid).

Removed

• #362 Removed epytope/BioMart variant annotation (epaa.py, EPYTOPE_VARIANT_PREDICTION, VARIANT_SPLIT) and params --genome_reference, --fasta_output, --split_by_variants*. Variant input is now raw somatic VCF (VEP is run in-pipeline) (@axelwalter)
• #379 Removed --netmhc_system and support for the macOS NetMHCpan and NetMHCIIpan tarballs, only the Linux tarballs are supported (@jonasscheid).