OncoRef 1.28.0: every protein resolves, with a coding universe, cell-type tiers and sequence rules

637 CTA / 18,278 Unsafe / 689 Other representative proteins across 19,604 proteins. No human protein is unassessed. Other includes 601 Nonreproductive, 34 Held, 28 Nested, 20 Contained and 6 Diffuse.

Rules and API · Full source report · Category funnel

Downstream semantics are unchanged: CTA is targeted, Unsafe blacklists any matching subsequence from CTAs, and everything else is ignored for targeting.

What changed

Change Effect
Universe Only proteins translated from a coding transcript or an established ncORF are classified (19,604). 259 groups are out of universe: 254 NMD-only or CDS-not-defined, 5 single-study Ribo-seq ORFs. They are listed with reasons in out-of-universe.csv.gz.
No Unassessed class The 647 Unassessed proteins resolve to Unsafe or Other. assessment_status records whether coverage was incomplete or a hold applied.
Coverage Only combined bulk (HPA consensus, HPA brain, GTEx v11) is required. The GTEx 52-context core and cell-cluster, immune, DICE and cornea panels never block a label.
Unsafe, higher bar Manageable-tier RNA above 128 is Unsafe evidence (12 proteins, mostly skin-barrier genes). A reviewed unresolved Protected-tissue finding is an essential unknown and is Unsafe (113 proteins).
Cell-type tiers One table (oncoref.cta_cell_tiers) feeds the RNA cascade and the rare-cell screen. Vital: neurons, glia, cardiomyocytes, alveolar types 1 and 2, endothelium, airway basal cells. Manageable: differentiated airway epithelium (ciliated, club, goblet, secretory, ionocytes), fibroblasts, bowel epithelium.
Rare vital-tissue screen A protein is Unsafe when at least 1% of a Vital cell type (brain, spinal cord, heart, lung) expresses it in CELLxGENE normal summaries. Evaluated for every protein; 117 are Unsafe only because of it. No CTA is flagged.
Contained A would-be CTA with at least 95% of its 9-mers inside Unsafe proteins is Other:Contained: its peptides are blacklisted. 20 proteins.
Nested A reference-CDS microprotein (under 50 residues) nested in a same-strand protein-coding gene, without ncORF evidence, is Other:Nested. 28 proteins.
Held Target gates pass, but a marginal or unresolved normal signal blocks CTA without establishing Unsafe: 34 proteins, including MAGEA3.

Unsafe reconciles exactly: 18,062 previous, minus 14 that left the universe, minus 9 released because differentiated airway cells moved to Manageable, plus 239 new (110 reviewed holds, 114 rare vital-tissue, 3 both, 12 Manageable above 128), gives 18,278.

CTA membership: 701 to 637

80 previous CTAs left and 16 entered. - 67 became Unsafe through the rare vital-tissue screen alone, for example CSAG1 (midbrain dopaminergic neurons 1.9%), PLAC1 (cerebellar granule cells 1.9%), DPPA2 (airway basal cells 1.7%), SOX30, SMC1B, TDRD1, TPTE, GDF9, MORC1 and HORMAD2. - 13 became Other:Contained: CGB2, CSH2, CT45A3, DAZ2, DAZ3, GAGE2E, PRAMEF27, RIMBP3C, TAF11L9, TBC1D3G/H/K and ZC3H11C. Most multi-copy families (CT45, GAGE, PRAMEF, POTE) stay CTA at the 95% floor; their usable subsequences are left for review. - PRAME, MAGEA4, MAGEA1, MAGEC2, NY-ESO-1, CTAG2, SSX2, CT83, XAGE1A/B, TBXT and GAGE10 remain CTA.

The 16 new CTAs include readthrough chimeras (ATF7-NPFF, EPPIN-WFDC6, SLCO1B3-SLCO1B7, TGIF2-RAB5IF) and unnamed loci, plus MESP2, MSGN1 and TSSK2.

Known limitations

  • Several new CTAs rest on weak or embryonic-only reproductive RNA (for example ENSG00000258417, 0.0 nTPM adult reproductive) and readthrough products share RNA with their parent genes. They pass every gate and are reported as such; review is advised.
  • The tumor gate remains one measured sample above 0, and the reproductive route accepts embryonic RNA of 5 TPM or more, so most proteins pass both.
  • The rare-cell screen reports detection frequency among cells, not abundance, and can reflect ambient RNA, doublets or paralog multimapping; samples cannot be restricted to adults. It flags 15,888 proteins, most already Unsafe. Whether a cell type counts as Vital or Manageable is a policy judgement. cta_rare_cell.UNSAFE = False makes it a hold.
  • MAGEA12 is Unsafe through its reviewed clinical exclusion and also now through the screen (midbrain dopaminergic neurons 1.5%). Research prioritization, not clinical safety.

Downstream adoption

from oncoref import cta_selection

results = cta_selection.results()        # category is never null for human
profiles = cta_selection.profiles()      # adds rare_cell_* and unsafe_sequence_overlap_pct
screen = cta_selection.rare_cell_screen()  # includes tier
roles = cta_selection.sequence_dispositions()

Update code that handles null human categories or the Unassessed label. New subtype values are Held, Contained and Nested. Proteins outside the coding or established-ncORF universe no longer appear in results().