GitHub
11/01/2026, 6:43 AMgene_start and gene_end in transcript cache files; the published cache predates those columns, so 0.11.0 stops on it with transcript table 'transcript' is missing required column gene_start.
This replaces only data/genomics/homo_sapiens/vepyr/cache.tar.gz (473,097 bytes; MD5 20b9da80db39a86054331589b55d9f4d). The rebuilt transcript parquet has both columns as int64, with the same archive members and entity manifests.
Verified with the PyPI wheels on the published input.vcf.gz and reference.fa.gz (1,000 variants). The expected record-body MD5 is that of Ensembl VEP 116, `1d4a92b815eb6193e6f4546f4ab35978`:
| vepyr | cache.tar.gz | Result |
| ------------------------------------------------ | ------------ | -------------------------------------- |
| 0.11.0 | this PR | MD5 reproduced |
| 0.10.0 (the vepyr/annotate module's current pin) | this PR | MD5 reproduced |
| 0.10.0 | published | MD5 reproduced |
| 0.11.0 | published | fails on the missing gene_start column |
The module's pinned 0.10.0 therefore keeps working with the refreshed archive, and its output is unchanged, so this can merge ahead of the module bump to 0.11.0.
No other test-data files changed.
🤖 Generated with Claude Code
https://claude.ai/code/session_01W815XL29mth9pwsSzTGFNU
nf-core/test-datasetsGitHub
11/20/2026, 9:00 AMGitHub
11/20/2026, 11:38 AM