GitHub
09/21/2026, 8:24 AMchr4, chr19, chrM. The fixture added in #2283 used Ensembl names, so a pipeline that has to rename germline contigs before matching them against somatic calls has that step exercised only as a no-op.
That is not hypothetical. In nf-core/epitopeprediction#362 the somatic calls were renamed and the germline ones were not, so no germline record ever matched a somatic window and the whole feature was a silent no-op for every real user. Both fixtures being Ensembl-named is exactly why CI stayed green; a real exome pair is what caught it.
Change
Only the contig names, in the two ##contig headers and the two records. Same variants at the same positions, nothing else touched:
-##contig=<ID=4,length=190214555>
+##contig=<ID=chr4,length=190214555>
-4 1801844 . C G 912.5 PASS ...
+chr4 1801844 . C G 912.5 PASS ...
Because renaming is meant to be transparent, a pipeline that does it correctly produces exactly the results it produced before, so the test using this file needs no new expectations and any difference is a rename failure.
Why not a second, parallel fixture
That was #2284, now closed. Adding UCSC-named copies alongside the Ensembl ones doubles the files and needs a new profile and test for a property the existing test can assert for free. Leaving the somatic fixture Ensembl-named is also deliberate: the mixed case, one file renamed and the other not, is the one that actually breaks.
🤖 Generated with Claude Code
nf-core/test-datasetsGitHub
09/21/2026, 8:59 AMGitHub
09/21/2026, 8:59 AM