OncoRef 1.33.0: CTA paper gaps, concerns and localization

The CTA report now evaluates all 33 proteins without qualifying paper coverage, summarizes the 29 flagged reviews, and provides filterable localization evidence for all 602 CTAs.

The 33 gaps span duplicated regulatory families, germ-cell structure, reproductive secretory families, maternal/oocyte programs and poorly characterized loci. All pass the existing coding-reference rule. Eighteen enter through developmental expression alone. Their median highest tumor RNA is 23.5 TPM, compared with 34.1 TPM for the 569 paper-covered proteins. Missing papers alone do not exclude a candidate; five have specific normal-cell questions requiring adjudication.

The concern table separates direct normal protein, physiological compartments, normal RNA, unresolved copies, plasma affinity assays, coding conflicts, proteoform attribution and limited characterization. Five unnamed loci have little literature; they do not have the experimental coding challenge reported for FRG2/FRG2B.

Localization combines a checksum-pinned UniProt 2026_03 extract with scoped primary-paper annotations. Fifty-six protein records have secretion or extracellular-matrix annotations, 37 have cell-surface or integral plasma-membrane annotations, and CRIPTO3 has extracellular detection with its route unresolved. ZP4 occurs in the first two categories. Shared-entry and partial-member annotations are separate; source evidence and model scope are retained. These counts include inferred annotations, not only experiments.

New APIs are cta_literature.gap_assessments(), cta_literature.concerns() and cta_literature.localization(). Each accepts an optional protein label or Ensembl gene ID. CSV downloads and a comparison figure accompany the tables. Candidate detail panels link to the localization evidence.

The classification snapshot remains 1.31.0: 602 CTA, 18,255 Unsafe and 1,006 Other. This release adds evidence and interpretation without changing the classification, expression datasets or paper-coverage calculation.

Validation covers all 33 gaps and 29 concerns, exact locus mapping, shared annotations, organelle-versus-surface distinctions, evidence types, report filters, offline reproduction, strict documentation builds and figure inspection.