Normal-tissue function the screen may miss: a literature sweep of all 902 candidates
The largest gap is tissue coverage, not thresholds. 242 of the 902 Target candidates belong to gene families whose normal site of action is a tissue the expression panels do not sample: nose, tongue, hair follicle and mucosal surfaces.
This is a screening aid produced by automated literature search. No category was changed. Every finding needs a human read of its source before it is used.
How it was done
37 research agents each took 25 candidates and searched the literature and curated databases for normal adult non-reproductive expression or function. A second agent then tried to refute each flagged finding against its cited source. 901 of 902 labels were covered (FSCB was skipped). 343 had essentially no literature.
| Outcome after the refutation pass | Findings |
|---|---|
| Confirmed: human, adult, normal tissue, gene-specific assay | 53 (51 proteins) |
| Confirmed with a limit (not adult human, or not gene-specific) | 11 |
| Weakened: mouse only, fetal, family-level or paralog-ambiguous | 184 |
| Refuted | 8 |
Every finding, verdict, source and reason.
1. Whole families sit in tissues the panels do not sample
| Family | Candidates | Normal site | In standard |
In strict |
|---|---|---|---|---|
| Olfactory receptors (OR) | 164 | Olfactory epithelium | 7 | 1 |
| Hair keratins and keratin-associated proteins (KRT3x, KRTAP) | 41 | Hair follicle | 0 | 0 |
| Beta-defensins (DEFB) | 22 | Epididymis; some in airway, gut and oral mucosa, some inducible | 2 | 1 |
| Taste receptors (TAS1R, TAS2R) | 9 | Taste buds; reported in gut and airway | 0 | 0 |
| Trace-amine and vomeronasal receptors (TAAR, VN1R) | 4 | Olfactory epithelium | 0 | 0 |
| Late cornified envelope (LCE) | 2 | Epidermis | 0 | 0 |
Most are Cancer Antigen only. They pass because bulk RNA of these tissues is absent
from the panels, so their normal maximum is near zero and a low tumor value clears it.
The paper layer already removes nearly all of them: standard keeps 9 of the 242.
The catalog itself does not.
2. Confirmed findings, by where the protein acts
| Where | Proteins |
|---|---|
| Cartilage, bone and muscle | CCN6, LRRC52 |
| Endocrine and metabolic | DEFB118, GALR3, GFRA4, KLF14, LALBA |
| Hair follicle | KRT37, KRT39, KRTAP10-6, KRTAP10-8, KRTAP12-4, KRTAP24-1, KRTAP29-1, KRTAP5-3, KRTAP5-4, KRTAP5-5, KRTAP5-7, KRTAP5-9 |
| Immune and blood | DEFB104A/B, DEFB130A/B, NLRP11, PRKACG, SCGB1D4, ZNRF4 |
| Kidney, gut, lung and other | DEFB106A/B, SPANXN3, SULT6B1 |
| Nose and olfactory | BPIFB3, OBP2A, SCGB1C1 |
| Sensory ganglia and brain | ADAM30, BECN2, GALP, MRGPRX1, SRY |
| Skin and sweat gland | AADACL3, ASCL4, OR51A7, PRAME, TCHHL1 |
| Taste and tongue | BPIFB6, RTP2, TAS2R1, TAS2R13, TAS2R41, TAS2R50, TAS2R60, TAS2R7 |
| Tooth | DSPP, PANX3 |
3. Confirmed findings that survive standard
| Protein | Where | Finding | Source | In strict |
|---|---|---|---|---|
| LRRC52 | skeletal muscle | Human: UniProt curates LRRC52 tissue specificity as 'mainly expressed in testis and skeletal muscle', citing Yan and Aldrich 2012 (PMID 22547800). The abstract of that paper, which I read, names only testis, so the skeletal muscle statement rests on the UniProt annotation of the paper's tissue panel. | PMID 22547800, full text PMC3356614, Fig. 4 and Discussion | yes |
| NLRP11 | blood / immune system | In human macrophages NLRP11 binds ASC and NLRP3 and is required for NLRP3 inflammasome assembly; NLRP11 deletion blocks NLRP3 activation and siRNA knockdown in primary human macrophages reduces IL-1beta release (Nat Immunol 2022). The gene is absent in mice. | PMID 35624206 (PMC9174058; DOI 10.1038/s41590-022-01220-3) | yes |
| PRKACG | blood / bone marrow | Human: UniProt P22612 curates germline PRKACG variants (Ile74Met) as the cause of Bleeding disorder, platelet-type, 19 (MIM 616176), a macrothrombocytopenia with platelet dysfunction, citing Manchev et al., Blood 2014. This implies a functional role in the megakaryocyte/platelet lineage, although UniProt's tissue-specificity not | Manchev VT et al., Blood 2014;124(16):2554-63, PMID 25061177, PMCID PMC4199957 ( | yes |
| ZNRF4 | blood / innate immune system | ZNRF4 is an ER-anchored E3 ubiquitin ligase that degrades RIP2 and negatively regulates NOD2 signalling: Znrf4-knockdown monocytes showed sustained NF-kB activation, and Znrf4-knockdown mice had reduced NOD2-induced tolerance to MDP and cleared Listeria monocytogenes better. The abstract does not state the species of the monocyt | PMID 28656966 (full text PMC5493756) | yes |
| SPANXN3 | several nongametogenic adult tissues (not named in abstract) | Human: SPANX-N family reported as expressed at a low level in several nongametogenic adult tissues, in contrast to SPANX-A/D. Family-level statement from the abstract; member-specific tissues not verified. | PMID 17406683 full text, Table 2 (PMC1831492) | no |
| ADAM30 | brain | Human: Letronne et al. 2016 (EBioMedicine, PMID 27333034) report from brain transcriptomics that ADAM30 expression is about 50% lower in Alzheimer's disease brains and inversely correlates with amyloid load, which implies measurable ADAM30 mRNA in non-diseased adult brain. The functional work (cathepsin D activation, APP lysosom | https://pmc.ncbi.nlm.nih.gov/articles/PMC4972530/ (PMID 27333034) | no |
| PRAME | skin | Human: nuclear PRAME staining in non-neoplastic skin is usually absent or limited to a few scattered melanocytes, most often in actinically damaged skin; the authors also note occasional weak nuclear staining of stromal cells in scars. This fits the catalog's existing dismissed rare-cell hit in eye melanocytes and hints at damag | PMID 34023098 (https://pmc.ncbi.nlm.nih.gov/articles/PMC8152939/) | yes |
| PANX3 | tooth (dental pulp) | Human dental pulp tissue and odontoblast-like cells express PANX3 mRNA and protein (RT-PCR and immunofluorescence), where it mediates ATP release on cold stimulus. | PMID 26263540 | no |
PRAME's finding is scattered nuclear staining in a few melanocytes of normal skin, mostly sun-damaged. It is recorded here for completeness and changes nothing.
4. A sequence gap: read-through proteins
Three candidates are read-through transcripts whose protein is mostly a ubiquitous parent protein: ATF7-NPFF (identical to ATF7 over 411 of 463 residues), ENSG00000273155 (the first 118 residues of MRPL30) and ENSG00000267618 (RAD51D and RFFL). The catalog's containment rule did measure them: 88.6%, 89.4% and 90.8% of their 9-mers lie in Unsafe proteins, just under the 95% cutoff, so all three are CTAs.
What was done with it
- A curation layer now carries these findings into the filters. Each confirmed
finding is assigned a tissue compartment and that compartment's tier, with its
citation kept.
normal_function="no_vital"(instandard) and"no_protected"(instrict) act on it. Layer and tiers. - Family rules extend three findings to whole families: olfactory receptors (olfactory epithelium, Protected), bitter taste receptors (taste buds, Protected) and hair keratins (hair follicle, Disposable). Trace-amine receptors, sweet and umami receptors and late cornified envelope proteins are recorded without acting, because the evidence is by similarity or not gene-specific.
- Read-through loci are removed by
readthrough="exclude"in every mode butall. - The catalog's categories are unchanged. Tiers for seven tissues the hazard registry
had no call for are proposals in
oncoref/data/target-curation.yaml, with references, and are not yet part of the registry.