Trim comments and changelog entry for the allele-label fix

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Author: jonasscheid

CHANGELOG.md

@@ -15,7 +15,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### `Fixed`
 
-- [#359](https://github.com/nf-core/epitopeprediction/pull/359) Fixed shuffled NetMHCpan/NetMHCIIpan allele labels in the merged/`best_allele` output by reading allele names from the xls header instead of assuming `sorted(meta.alleles)` order (fixes [#358](https://github.com/nf-core/epitopeprediction/issues/358)) ([@jonasscheid](https://github.com/jonasscheid/))
+- [#359](https://github.com/nf-core/epitopeprediction/pull/359) Fixed shuffled NetMHCpan/NetMHCIIpan allele labels by reading allele names from the xls header (fixes [#358](https://github.com/nf-core/epitopeprediction/issues/358)) ([@jonasscheid](https://github.com/jonasscheid/))
 - [#352](https://github.com/nf-core/epitopeprediction/pull/352) Accept 3- and 4-field HLA typings by truncating to 2 fields via mhcgnomes (fixes [#350](https://github.com/nf-core/epitopeprediction/issues/350)) ([@jonasscheid](https://github.com/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](https://github.com/jonasscheid/))
 - [#349](https://github.com/nf-core/epitopeprediction/pull/349) Fixed MultiQC stats mismatch: use per-peptide binder counts, fix unsupported count formula, fix netmhcpan multi-allele column parsing ([@jonasscheid](https://github.com/jonasscheid/))

modules/local/merge_predictions/templates/merge_predictions.py

@@ -195,11 +195,7 @@ def _format_mhcnuggets_prediction(self) -> pd.DataFrame:
         return df
 
     def _format_netmhcpan_prediction(self) -> pd.DataFrame:
-        # Map output-column index to allele from the xls header row (one name per allele block), in
-        # NetMHCpan's output order. Reconstructing the order from sorted(meta.alleles) mislabels
-        # alleles whenever that sort diverges from NetMHCpan's — e.g. a stray space in a samplesheet
-        # allele ("; C*07:04") makes it sort first — leaving ranks correct but labels shuffled (#358).
-        # These names are normalized by mhcgnomes for all predictors in main().
+        # Map column index to allele from the xls header (output order), not sorted(meta.alleles) (#358)
         with open(self.file_path) as f:
             header_alleles = [allele.strip() for allele in f.readline().split('\t') if allele.strip()]
         alleles_dict = dict(enumerate(header_alleles))
@@ -245,11 +241,7 @@ def _format_netmhciipan_prediction(self) -> pd.DataFrame:
         NetMHCIIpan 4.3 xls uses `Rank` for the EL percentile and `Rank_BA` for BA percentile;
         multi-allele files are handled by pandas auto-suffixing duplicate columns as `.1`, `.2` …
         """
-        # Map output-column index to allele from the xls header row (one name per allele block).
-        # NetMHCIIpan re-sorts alleles in its own name format, which diverges from the pipeline's
-        # mhcgnomes sort (e.g. DPB1*10:01 vs DPB1*107:01), so reconstructing the order from
-        # sorted(meta.alleles) mislabels alleles while leaving ranks correct (#358). These names are
-        # normalized by mhcgnomes for all predictors in main().
+        # Map column index to allele from the xls header (output order), not sorted(meta.alleles) (#358)
         with open(self.file_path) as f:
             header_alleles = [allele.strip() for allele in f.readline().split('\t') if allele.strip()]
         alleles_dict = dict(enumerate(header_alleles))