Current CTA and oncofetal selection funnel

The complete protein-group funnel selects 343 CTA, with 17,824 Unsafe and 1,696 Neither among 19,863 protein equivalence groups. These represent 20,097 reference gene loci; 358 reference loci belong to selected groups. Before grouping, the same evidence policy assigns 359 CTA / 17,906 Unsafe / 1,832 Neither at the gene-locus level. The two denominators are intentionally reported separately.

Uniform rules

1. Reproductive HPA RNA ≥5 nTPM and greater than nonpermitted adult bulk, or a supported Tyser embryonic lineage mean ≥5 TPM (≥20 cells, ≥5 positive cells, ≥10% at TPM≥1). No gene-specific literature admission route.
2. At least one qualifying tumor sample ≥2 native TPM/pTPM. The complete compatible OncoRef matrix union is used; same-sample group sums precede maxima.
3. Require bulk/HPA-cluster/HPA-immune/Monaco coverage. DICE and cornea are optional for coverage; observed expression still contributes.
4. Apply the linked 2×2 normal-RNA/IHC grid. Above the candidate ceiling is a hold; ≥2× the ceiling supplies Unsafe evidence. Preserve exact anatomy and assay units.
5. Retain credible normal-protein and clinical-safety exclusions and externally reviewed findings. An antibody concern disables negative-IHC threshold boosts; it does not impose an additional gene-wide veto after the stricter RNA screen. Context-specific positive findings remain active. The current external collection supplies 0 fully context-validated observations and no automatic negative overrides.

CTA means the research selection rules pass, Unsafe means a research-policy exclusion, and Neither means an admission, coverage or unresolved-evidence hold. Every layer evaluates every entity; later Unsafe evidence can supersede an earlier Neither. Released annotations and forbidden sequence references are not overwritten.

Complete cascade

layer	new_candidate_removals	new_Unsafe	CTA_eligible_so_far	Unsafe	Neither
Reproductive / embryonic association	7724	0	12139	0	7724
Tumor sample ≥2 TPM/pTPM	7	0	12132	0	7731
HPA normal protein	9005	10391	3127	10391	6345
Required RNA coverage	87	0	3040	10391	6432
Normal bulk	2404	6354	636	16745	2482
HPA tissue clusters	247	898	389	17643	1831
HPA immune	3	44	386	17687	1790
Monaco immune	10	40	376	17727	1760
DICE immune	28	83	348	17810	1705
Adult cornea	4	10	344	17820	1699
Accepted normal-protein evidence	1	3	343	17823	1697
Clinical safety exclusion	0	1	343	17824	1696
External validated normal protein	0	0	343	17824	1696

Sequential counts, independent overlapping effects, gene-locus counts.

Named antigens and known CTA tracking

protein_group_label	category	exclusion_reasons
PAGE1	CTA	All required gates pass
SMC1B	CTA	All required gates pass
CTAG2	Unsafe	Normal bulk; HPA tissue clusters
MAGEA4	CTA	All required gates pass
TBXT	CTA	All required gates pass
PRAME	Unsafe	Normal bulk; HPA tissue clusters; Accepted normal-protein evidence
CT83	CTA	All required gates pass
NY-ESO-1	CTA	All required gates pass
SSX2/B	CTA	All required gates pass
XAGE1A/B	CTA	All required gates pass

Current decisions for mapped CTA/source-list genes retain each gene's source lists, locus decision, group decision and all exclusion reasons. The current funnel evaluates the complete reference universe; previous CTA labels are comparison annotations and do not define its candidate set or admission rules. Imported lists include broad nomination lists; membership is not a claim that every entry is an established CTA. Other MAGEs receive ordinary evaluation, not control status.

Source category counts, source overlap matrix. HPA reproductive and Tyser embryonic measurements are biological nomination routes; Tyser is one early embryo, not a comprehensive fetal atlas or validated oncofetal target list. Adult normal-reference coverage does not establish absence in pediatric tissues.

Identity, aggregation and evidence limits

• Every gene result carries its canonical ENSG, registered historical/alternate IDs, synonyms and identity-catalog hash. Ambiguous names and contradictory IDs are quarantined rather than assigned to the first match.
• Protein equivalence groups describe identical chosen reference proteins. Their IDs hash member identity; they are not sequence hashes or all-isoform equivalence. Complete member identity bundles and the pinned group-registry hash are retained.
• CTAG1A/B and XAGE1A/B are summed only within matched samples or tissue contexts. Bulk HPA and GTEx components are summed separately before taking a source maximum. Aliases of one locus are never added as extra expression.
• DICE lacks donor pairing across genes. Ranked-sum bounds retain this uncertainty; a possible upper exceedance causes review, not invented Unsafe evidence. Missing optional DICE coverage alone does not hold a group.
• Embryonic lineage means and corneal context means are additive. Detection-cell union bounds remain explicit where raw cell unions are unavailable. Embryonic admission and strong corneal evidence use the conservative supported bound.
• New external findings retain categorical staining and numeric scores, assay, specimen state, reagent lineage, missing donor counts and exact-context limitations. CT83's positive tumor-adjacent gastric evidence is flagged; adult disc findings remain attached to TBXT. Neither is silently recoded as healthy normal absence.
• TBXT now passes the strict RNA fallback, with the disputed HPA reagent and adult disc evidence retained as review flags. This does not claim an experimentally validated negative in every adult compartment or clinical safety.
• PRAME has substantial clinical target-specific evidence from IMA203. Its expression-screen exclusion must not be read as demonstrated clinical unsafety. See the PRAME, CTAG2 and TBXT evidence assessment.

Downloads and reproduction

Selected CTA groups, Unsafe, Neither, all group decisions, all locus decisions, locus/group crosswalk, changes caused by aggregation, named controls.

Validation record records the checks against this report's manifest.

Run the expanded tumor and normal builders, then the expanded funnel, followed by scripts/cta_protein_group_normal.py and scripts/cta_harmonized_report.py with this audit directory. External evidence is rebuilt by scripts/collect_external_ihc.py and scripts/curate_external_ihc.py. Source receipts, input hashes and output hashes are in the manifest and component manifests. Every figure also has an editable SVG alongside its PNG.