UniProt proteome removal

Which gene reviews lose their UniProtKB entry

AI Gene Review · projects/PROTEOME_REMOVAL · 2026

Bottom line

  • UniProt listed ~58M TrEMBL entries from redundant and non-reference proteomes for removal by release 2026_02 (archived in UniParc).
  • A check of the 896 accessions in the repo on 2026-02-03 found 15 reviews on the list; none has been re-keyed or archived.
  • On 2026-09-26, 5 are deleted from UniProtKB and 10 are still live. The repo is now ~4,990 folders, so the scan needs re-running.

Why it matters

The 15 flagged reviews today

How the check was done

  • Used UniProt's explicit removal list (proteins_to_remove_from_UniProtKB.txt, 578 MB), not proteome membership, which proved unreliable.
  • Intersected with the repo's accessions using LC_ALL=C comm -12 on sorted lists: linear time, but locale-sensitive.
  • The just targets described on the project page are not in the current justfiles.

Status and next steps

  • ✅ Removal list checked against 896 accessions (Feb 2026): 15 hits.
  • ⬜ For the 5 deleted entries (merB, xdhB, fae1A, stx2A, Q1IFG0): find a retained accession or archive the review.
  • ⬜ Decide whether a UniParc UPI is an acceptable review key.
  • ⬜ Restore the check as a just target and re-run over all ~4,990 gene folders.

Read more: projects/PROTEOME_REMOVAL.md