CTA, Unsafe and Neither
589 CTA candidates, down from 740. 151 previous candidates move to Neither, not the forbidden background. The rule is at least 5 nTPM reproductive RNA when there is no positive IHC anywhere measured. All 21 currently retained clinical/requested benchmark genes remain CTA candidates.
Find any gene · Removed candidates and reasons · Gene sets · Protein downloads · Threshold comparison
The three sets
| Set | Meaning | Target? | Forbidden background? |
|---|---|---|---|
| CTA | Curated candidate passing the existing RNA/IHC/selectivity rules and the new admission floor | Yes | No |
| Unsafe | Positive IHC in Vital/Protected/Manageable tissue, above-ceiling normal RNA, or an independent reviewed safety hold | No | Yes |
| Neither | Low-expression, Variable, unresolved, uncurated or selectivity-only failures without decisive normal-expression exclusion | No | No |
These are research labels. CTA means eligible under this screen, not proven safe treatment; Unsafe means operationally forbidden under the policy, not demonstrated clinical toxicity. Neither is not evidence of protein absence.
| Category | Genes | Exact protein sequences |
|---|---|---|
| CTA | 589 | 1138 |
| Unsafe | 17405 | 125578 |
| Neither | 2103 | 3269 |
The gene universe is 20,097 canonical human protein-coding genes, not just the 2,540 source-nominated CTA candidates. All genes · CTA · Unsafe · Neither · Candidate funnel · Neither subgroups.
What changed
- 150 previously retained genes have reproductive RNA <5 nTPM and no measured IHC at all. Their RNA is also <5 nTPM in every measured tissue.
- IL5 adds one more removal: reproductive maximum 4.9 nTPM, measured IHC but no positive IHC anywhere. “No positive IHC” therefore removes 151, rather than the 150 estimated for “no IHC.”
- CTAG2 remains Neither / Variable, with its reviewed donor RNA and negative local IHC preserved. Negative IHC and RNA donors are different people; only CTAG2 has imported individual-donor RNA evidence.
- ZNF730 is Neither / Low: it fails CTA selectivity and has low RNA without positive protein evidence, but does not exceed its normal RNA ceilings. It is not automatically a forbidden protein.
- Positive normal IHC, including Low, and independent safety holds still take precedence. MAGEA12 stays Unsafe.
| Gene | Category | Detail | Reproductive RNA (nTPM) | IHC coverage |
|---|---|---|---|---|
| MAGEA4 | CTA | Pass | 21.5 | Positive somewhere |
| CTAG1B | CTA | Pass | 7.4 | Positive somewhere |
| PRAME | CTA | Pass | 102.2 | Positive somewhere |
| CT83 | CTA | Pass | 62.2 | Missing |
| XAGE1A | CTA | Pass | 45.7 | Missing |
| XAGE1B | CTA | Pass | 53.1 | Missing |
| MAGEA3 | CTA | Pass | 7.8 | Missing |
| CTAG1A | CTA | Pass | 8.1 | Positive somewhere |
| IL5 | Neither | Low | 4.9 | Negative where measured |
| CTAG2 | Neither | Variable | 10.1 | Positive somewhere |
| ZNF730 | Neither | Low | 4.9 | Missing |
| MAGEA12 | Unsafe | Safety | 2.7 | Missing |
The 5 nTPM value is an admission floor, not a replacement normal-tissue safety ceiling. A gene below 5 nTPM everywhere can still be Unsafe if, for example, brain RNA exceeds its applicable 2 nTPM baseline. Missing measurements never become zero. Low requires usable routine RNA coverage; missing/unreliable evidence remains explicitly Unassessed within Neither.
IHC counts describe curated tissue/cell-type annotations, not independent donor counts. Missing IHC and measured-negative IHC are reported separately.
The existing tissue rules remain
Each cell is maximum nTPM / nominal minimum selectivity. Relaxation requires reliable negative IHC across routine local cell types plus reliable positive reproductive IHC in the curated profile.
| Tissue | Baseline | Approved negative | Supported negative | Enhanced negative |
|---|---|---|---|---|
| Vital | 2 / 80% | 4 / 75% | 6 / 70% | 12 / 65% |
| Protected | 6 / 75% | 8 / 70% | 12 / 65% | 24 / 60% |
| Manageable | 10 / 70% | 20 / 65% | 30 / 60% | 60 / 55% |
CTA selection still requires N ≤ C, N ≤ R × (1−s)/s + 0.1, and N < R. A selectivity failure alone removes CTA eligibility; it does not create an Unsafe assignment. Reviewed Variable evidence is a separate reconsideration route, excluded from both target and forbidden sets unless competing protein/safety evidence wins. Prior ordered rules and donor evidence.
Protein sets and k-mer exclusions
Complete identical amino-acid sequences are assigned once, with every encoding gene/transcript/protein occurrence retained. Priority is Unsafe → Neither → CTA. Thus an Unsafe encoding gene makes the shared protein forbidden; an unresolved/Neither encoding gene blocks target selection without making that protein forbidden. XAGE1A/B remain separate loci while their identical proteins are grouped.
| Set | Protein inventory | FASTA |
|---|---|---|
| CTA | Proteins | Sequences |
| Unsafe | Proteins | Sequences |
| Neither | Proteins | Sequences |
The forbidden union is unsafe.fasta.gz plus additional-forbidden.fasta.gz. Neither, Variable, Low and Unassessed are not added. The additional file contains incomplete/ambiguous translation records only from Unsafe genes. The historical Normal + Unassessed exclusion recipe is not the current default.
Full proteins are disjoint; k-mers can still overlap. Subtract all forbidden k-mers from CTA k-mers and never join across record boundaries, ambiguous residues or stops. 34,361 translation records require review; 307 classified genes have no complete usable protein. Their gaps are preserved, not interpreted as absence. Gene RNA does not establish isoform-specific translation or peptide-HLA presentation.
Gene and protein counts differ. For example, MAGEA3 passes at the gene level, but its complete reference protein is also encoded by unresolved alternate-contig gene ENSG00000292186. That shared protein stays Neither pending identity review, as it did in the previous Unassessed sequence export. Matching the gene name alone does not resolve the locus. XAGE1A/B and CTAG1A/B shared proteins remain in CTA.
Machine-readable recipe · Additional forbidden records · All proteins · All occurrences · Translation review · Missing complete proteins · Unclassified sequence genes.
Provenance and API
Pinned human HPA v23 expression and Ensembl 115 / GRCh38.p14 complete-sequence inventory. Existing source tissue/cell mappings and antibody confidence are preserved; human thresholds are not transferred to dogs or mice. Original dated reports remain available. HPA sources and antibodies · Reference identity · Input/code/output checksums.
from oncoref import target_sets
genes = target_sets.results()
targets = target_sets.gene_ids("CTA")
forbidden = target_sets.gene_ids("Unsafe")
neither = target_sets.gene_ids("Neither")
Use target_sets for the current three-way sets. Older cta_screen results and normal-sequence exports retain their explicitly dated policies.