A Bounded Pre-Wet-Lab Evidence Case for the APM-Loader RNA Signal in Adult Cutaneous Melanoma

GTIM-ADULT-SKCM-APM-001 A Bounded Pre-Wet-Lab Evidence Case for the APM-Loader RNA Signal in Adult Cutaneous Melanoma Research White Paper + Supplementary Replication Protocol GTIM Adult Oncology Public Evidence Desk Case ID GTIM-ADULT-SKCM-APM-001 Disease context TCGA-SKCM — Skin cutaneous melanoma Route APM_TAP_LOADER_IMMUNE_VISIBILITY_ROUTE Decision GO — bounded pre-wet-lab priority Engine gtim_adult_oncology_public_evidence_desk_v0_7_8 / v0.7.8 Completion marker REPORT_COMPLETE_V0_7_8 Clinical boundary Research Decision Support only Document status Publication-ready white paper draft; remove Appendix B for journal submission Use boundary: this document is intended for research prioritization, partner review, repository release, or conference-style white paper circulation. It must not be presented as a biomarker validation, clinical test, treatment recommendation, or medical-device output. Suggested Citation GTIM Adult Oncology Public Evidence Desk. GTIM-ADULT-SKCM-APM-001: A bounded pre-wet-lab evidence case for the APM-loader RNA signal in adult cutaneous melanoma. Publication draft with supplementary replication protocol. Version 1.0, 2026. Document Map 1. Scientific Core: hypothesis, evidence result, interpretation, and decision boundary. 2. Methods Lock: score definition, endpoint, Cox model, and downgrade criteria. 3. Supplementary Protocol 1: independent survival replication protocol. 4. Appendix B: optional market-facing partner note; remove for journal submission. 5. References. Abstract Background: AI-generated and route-level oncology hypotheses are increasingly easy to produce, but costly to test experimentally. A pre-wet-lab evidence desk can reduce waste by determining whether a submitted hypothesis deserves a specific next evidence step. Objective: This document summarizes GTIM-ADULT-SKCM-APM-001, a bounded evidence case evaluating whether low B2M/TAP1/TAP2/TAPBP antigen-presentation loader RNA state is associated with worse overall survival in adult cutaneous melanoma. Methods: GTIM v0.7.8 resolved the hypothesis to TCGA-SKCM, applied a hard disease-context lock, constructed a route-scoped APM-loader RNA signal, and executed matched expression-survival analysis using Cox proportional hazards modeling. Result: The complete-evidence run aligned 456 cases with 220 events and produced HR/1SD = 0.678, p = 8.16e-06. Because the predictor is increasing APM-loader score, HR/1SD < 1 supports the submitted low-APM-worse direction. Conclusion: The case receives GO for bounded pre-wet-lab prioritization only. It does not prove surface MHC-I/pMHC, antigen-presentation function, immune killing, checkpoint response, treatment selection, or clinical actionability. The next required step is an independent matched adult melanoma survival replication using the locked score and direction rule. Keywords GTIM; melanoma; TCGA-SKCM; antigen presentation; APM-loader; B2M; TAP1; TAP2; TAPBP; MHC-I; pre-wet-lab evidence; survival replication; research decision support. 1. Executive Summary GTIM v0.7.8 evaluated the submitted hypothesis: In adult melanoma, low B2M/TAP1/TAP2/TAPBP antigen-presentation loader RNA state is associated with worse overall survival. The engine resolved the disease context to TCGA-SKCM — Skin cutaneous melanoma, executed a matched public expression-survival analysis, and produced a GO decision under a strict pre-wet-lab boundary. Dataset/context TCGA-SKCM — Skin cutaneous melanoma Primary route APM_TAP_LOADER_IMMUNE_VISIBILITY_ROUTE Aligned cases n = 456 Events 220 Cox result HR/1SD = 0.678 P-value 8.16e-06 Decision GO — bounded pre-wet-lab priority Decision reason MATCHED_PUBLIC_EXPRESSION_SURVIVAL_SUPPORTS_PRECLINICAL_PRIORITY Direction interpretation is essential: the Cox model uses increasing APM-loader score as the predictor. HR/1SD < 1 means higher APM-loader RNA state is associated with lower hazard. Therefore, the submitted low-APM-worse direction is supported. This result does not prove surface MHC-I expression, functional antigen presentation, immune killing, treatment response, or clinical utility. It supports only a bounded research decision: the route is worth the next matched evidence step before any wet-lab expansion. 2. Biological Problem and Rationale Melanoma is immunologically active, but immune visibility is not guaranteed. A tumor can carry antigens and still avoid CD8 T-cell recognition if antigen processing and MHC-I presentation are weakened. The antigen-presentation machinery includes components such as B2M, TAP1, TAP2, and TAPBP/tapasin, which support peptide loading and presentation through MHC class I. Loss or suppression of MHC-I antigen presentation is a recognized immune-evasion route in cancer. This makes the APM-loader state biologically plausible but not automatically decision-ready. GTIM’s role is narrower: determine whether a submitted route-level hypothesis has enough matched patient-level signal to justify the next evidence step. 3. Hypothesis and Claim Boundary Primary hypothesis Low B2M/TAP1/TAP2/TAPBP antigen-presentation loader RNA state is associated with worse overall survival in adult melanoma. Predictor used in Cox model Increasing APM-loader RNA score. Score definition Mean z-score of standardized log2 expression values for B2M, TAP1, TAP2, and TAPBP. Claim class Bulk RNA expression-survival association. Supported direction HR/1SD < 1 for increasing APM score supports the low-APM-worse wording. Claims explicitly not tested · Surface MHC-I or pMHC abundance. · Tumor-cell-specific attribution of the RNA signal. · Functional antigen presentation. · CD8 T-cell recognition or immune killing. · Checkpoint-inhibitor response. · Clinical stratification, diagnosis, treatment selection, or patient-specific actionability. 4. Evidence Source and Execution GTIM used the adult melanoma context TCGA-SKCM. TCGA is a public cancer genomics program that molecularly characterized more than 20,000 primary cancer and matched normal samples across 33 cancer types, making it a suitable source for exploratory route-scoped evidence checks. GTIM v0.7.8 executed a matched expression-survival analysis using route-sliced expression data, endpoint parsing, Cox proportional hazards modeling, and a hard context lock to prevent stale UI defaults from redirecting melanoma hypotheses to unrelated projects. v0.7.6 to v0.7.8 comparison Metric v0.7.6 v0.7.8 Interpretation Aligned cases 248 456 Larger matched evidence base. Events 131 220 Improved statistical information. HR/1SD 0.625 0.678 Same protective direction; effect moved modestly toward 1. P-value 0.000168 8.16e-06 More statistically stable signal with larger sample. Decision GO GO Decision preserved under complete-evidence run. The correct methodological conclusion is that v0.7.8 increased coverage and statistical stability. The effect size did not become biologically stronger; HR moved modestly toward 1 while the larger event set increased precision. 5. Decision and Interpretation GTIM decision GO — bounded pre-wet-lab priority. Meaning The hypothesis is strong enough to justify one next matched evidence object. Not meaning Not a biomarker, not a therapy-response predictor, not a clinical stratifier, not a medical-device output. Main red flag Bulk RNA does not prove surface MHC-I/pMHC, antigen-presentation function, or immune killing. One next evidence item Independent matched adult melanoma evidence object using the same APM-loader score and survival direction rule. Invalidate rule Downgrade if an adequately powered independent adult melanoma cohort reverses the direction or if the signal disappears after predeclared context adjustment. The route genes may be expressed in tumor cells, immune cells, stromal components, or mixed tissue fractions. A higher APM-loader RNA score may reflect tumor-intrinsic antigen presentation, immune-rich microenvironment, interferon activation, or a combination of these. Therefore, the GO decision must remain bounded. 6. Evidence Ladder Tier 1 — Independent survival replication Independent adult melanoma RNA-survival cohort; same B2M/TAP1/TAP2/TAPBP score; same direction rule; Cox or endpoint-equivalent model; predeclared covariates. Tier 2 — Cell-source attribution Single-cell RNA-seq, spatial transcriptomics, sorted tumor-cell data, tumor purity adjustment, immune-infiltration proxy. Tier 3 — Surface validation Surface MHC-I/B2M/HLA protein evidence by flow cytometry, IHC, multiplex IF, or mass cytometry. Tier 4 — Functional validation Antigen-presentation assay, T-cell recognition assay, perturbation/rescue of TAP1/TAP2/B2M/TAPBP, immune-killing readout. Tier 5 — Translational extension Checkpoint-response association only after tumor-cell and functional evidence pass. 7. Why This Case Matters for GTIM This case is strategically important because it demonstrates that GTIM can discriminate between different claim failure modes and bounded positive evidence. In the same Adult Oncology validation battery, ATR/ceralasertib in adult AML was blocked because a drug-sensitivity claim requires functional response evidence, while ERN1/XBP1 UPR in adult AML was blocked because matched expression-survival analysis did not support the claimed worse-survival direction. The APM/TAP-loader melanoma hypothesis passed because matched public expression-survival evidence supported the direction in a complete-evidence run. Negative control 1 ATR / Adult AML — NO-GO due to missing functional drug-response