Fixed

• Batch binding affinity predictions: Split large FASTA files into batches of 500 sequences before sending to NetMHCpan/NetMHCIIpan, preventing indefinite hangs on large inputs (e.g., 3,700+ alternative splicing events). Also fixed as_completed loop running outside the executor context and added detailed progress logging. (#58)
• Fix malformed params and string comparisons: Fixed align.smk params that passed a literal string instead of the config MAPQ value, and replaced is not "" with != across compile.py and merge_counttables.py (future Python 3.12 SyntaxError). (#61, #72)

Refactored

• Standardize log paths and fix missing log redirects: Unified log directory structure across all 138 rules (logs/ref/, logs/download/, logs/{sample}/{stage}/), added missing > {log} 2>&1 redirects to ~40 shell rules, and fixed several log path bugs (typos, duplicates, SE/PE mismatches). (#60)

Fix

• Improved error logging and output capture for sample processing steps.
• Optimized the workflow by sorting input files by QNAME before splitting
  in mhc-I genotyping, and removing the indexing step

Fixes

• Fixed bug in prioritization of MHC-II variants. Added wrong perl path to IEDB tools
• Enhanced validation to skip malformed amino acid change entries

Chores

• Updated workflow environment configuration with additional dependencies

Mostly addresses issues defined in #54

Fix

• added 10 retries to curl calls to avoid errors when downloading files (VEP cache, plugins)
• prioritization of MHC-II variants uses now the standalone version of IEDB
• explicitely chop HLA alleles to first two fields (e.g., HLA-A02:01:01 becomes HLA-A02:01)
• changed mhc-II refset to match IEDBs binding affinity prediction
• shell=True caused interactive python session instead of subprocess (when predicting binding affinities MHC-II))

Changed

• code polishing
• install instructions for HLA-HD in README

Fix

• Moved script for preparing hla input to workflow/scripts/genotyping
• Added missing pipe to log file in rule hlatyping_mhcII
• Added wrong input file to filtering of mhcII reads on SE

### Fix

• Fixed parameter in .tests/integration/config_basic/config.yaml (align) to match general config

Fix

• Removed fixed versions for bwa (0.7.19) and samtools (1.21) to allow installation of the latest versions.
• Updated the VEP wrapper version from "v1.31.1/bio/vep/annotate" to "v5.9.0/bio/vep/annotate".
• Modified functions for file type handling: adjusted indentation in get_preproc_input, updated get_input_filter_reads_mhcII_PE to handle BAM files directly, renamed get_aligned_reads to get_aligned_reads_featurecounts, added new function get_output_hlatyping_mhcII, and updated rule input parameters accordingly.

Fix

• Changed protocol for HLA alleles reference list to https (rule get_hla_info)
• Fixed path to input files in finalize_mhcII_input.py (which caused error when using paired-end reads)
• Fixed bug in quantification/featurecounts - supported regardless of input type (PE or SE): added wrapper script
• replaced MHC-II binding affinity prediction with IEDB API
• Manual Dockerfile replaces containerized version

Features

• Added sequence similarity filter for MHC-I
  • self-similarity (using kernel similarity)
  • pathogen similarity (BLAST against pathogen-derived epitopes from IEDB)
  • proteome similarity (BLAST against human proteome)
• Prioritization of neoantigens is now done separately for each variant type (speeds up the process)
• Update to recent version of ScanExitron (Thanks to @dolittle007)
  • this version updated to recent version of regtools (v0.5.0) - which is available on Conda
  • Singularity/Docker is not necessary anymore
  • Added option to use strand information in exitron calling
• ScanNeo2 now uses conda environments for all tools (ditched Singularity/Docker)

Fix

• renamed similarity fields for pathogen and protein to more descriptive names

Fix

• Fixed missing alleles in HLA alleles reference list - #34