CTA screening rules

For the current selection and protein backgrounds, use the CTA, Unsafe and Neither sets. The v1.15 policy below remains the dated 740-candidate input to that new classification.

The current report applies a simple ordered screen to the pinned human HPA v23 candidate universe. Its categories are Retained, Variable, Excluded and Unassessed. Variable is excluded from selection, with a concrete reason to reconsider after new evidence.

Rules

  1. Require reproductive expression, usable anatomy/identity/coverage and no independent curation or safety exclusion.
  2. Positive IHC (including Low) in a Vital, Protected or Manageable compartment excludes. Reproductive or Disposable staining does not trigger this veto.
  3. Relax RNA limits only when local normal IHC is reliably negative in all routine measured cell contexts and the same curated gene profile has reliable positive reproductive staining. Otherwise use baseline limits.
  4. Require every normal tissue to meet the absolute ceiling, reproductive selectivity and strict reproductive-higher comparison. The previous numerical ladder is tightened by five percentage points; absolute ceilings and the fixed 0.1 nTPM allowance remain. The report displays the entire matrix.
  5. Apply those same inclusion rules to each reviewed individual Vital-tissue RNA donor. A failing donor excludes even if the tissue mean passes.
  6. Label that exclusion Variable when a different donor in the same cohort reports zero RNA and the affected tissue has complete reliable negative IHC, with no competing protein/safety exclusion or unresolved quality flag.

There is no separate fixed high-donor RNA threshold. Zero is the source's reported estimate, not proof of absence. Missing data are never zero. Cohorts, source releases, genes, tissues and units are kept separate. Technical duplicate specimens cannot manufacture two independent donors.

Evidence scope

The positive reproductive control currently comes from HPA's curated gene profile. These exports do not establish the same antibody, experimental run or donor for every comparison. The gate is an interpretable research convention; it does not calibrate antibody sensitivity. Enhanced validation based partly on RNA is not an independent RNA replication. Primary staining and curated interpretation must be distinguished when presumed off-target signal was disregarded. See HPA's methods and application-specific antibody validation.

Currently only CTAG2 heart has imported individual-donor RNA. Its four reviewed v23 values (13.5, 0, 0, 7.1 nTPM) are a structured extraction of the existing source-cited donor review. Current image metadata confirm that the negative heart IHC donors are different individuals. These live IHC records are supplementary provenance, not replacements for the pinned v23 annotations. All other genes' donor profiles remain Not assessed, including retained genes.

To reconsider Variable: determine the source of donor variation, locate the expressing cells and relevant isoforms, obtain matched RNA/protein evidence in RNA-positive and RNA-zero specimens, and validate any proposed subgroup before an explicit new review. An RNA-zero donor is not automatically an eligible patient. Variable is not a clinical diagnosis or proof of treatment toxicity.

Python

from oncoref import cta_screen

decisions = cta_screen.results()
selected = cta_screen.retained_gene_names()
later = cta_screen.variable_gene_names()  # CTAG2; excluded from selected
selected_ids = cta_screen.retained_gene_ids()  # all 740 loci, including unnamed genes
later_ids = cta_screen.variable_gene_ids()    # disjoint from selected_ids
donors = cta_screen.reviewed_donor_rna()

# Evaluate source-shaped candidate/audit frames; all other exclusion gates remain.
decisions, tissues, donor_findings = cta_screen.screen_candidates(
    candidates, original_hpa_v23_tissue_audit, donors=donors
)

Retained means a point-screen pass on the reviewed evidence. No fitted donor model, antibody false-negative probability, peptide-HLA absence or clinical safety probability is implied. The earlier uncertainty interface remains available for independently fitted and validated model outputs. Symbol accessors omit loci without a current symbol; use the ID accessors to retain every locus without substituting a paralog.

The published core/extended CTA APIs retain their original policies and already exclude CTAG2. This new explicit selection also excludes CTAG2. Historical reports and sequence exports retain their dated provenance; this report does not generate new k-mer exclusion backgrounds. The tissue taxonomy supports other species, but human HPA nTPM limits are not transferred to dog/mouse measurements.