GALENOS 1-Auditor Biomedical Simulation, Open-Cohort & Regulatory Workbench
✓ 52/52 G1–G4 PASS ! S-SIGN: UNSIGNED (NULL) KR · 한국어
Wave 1 · KHIDI RFP 19-1 (p.104–106) · TRL 3 → 5 S-PUB ✓ PASS
Solo Reached: 89/100 · +CRO/Clinical: 96/100

Joint Genotype × HPO Likelihood & ACMG-28 Rare-Disease Diagnostic CDSS

Joint VCF x ACMG-28 x HPO Bayesian ranking reaches 100.00% Top-5 (97.92% Top-1) causal variant recall across 240 positive diagnostic cases (vs 56.67% noisy HPO-only and 9.18% post-2022 Phenopacket HPO-only), with 0.9814 AUROC across 6,011 ClinGen expert-classified variants.

2. Solver-Free Recomputed Benchmark Metrics

Target: >= 90.0% Causal Variant Top-5 Recall with Zero Self-Lookup Leakage
Causal Gene+Variant Top-5 Diagnostic Recall (Noisy HPO + VCF) (n = 240)
100.00 %
95% CI [98.42, 100.00]
Baseline Comparison (Genome-Wide HPO Resnik IC Phenotype-Only Ranker (4,900 Genes, Noisy HPO))
56.67 %
Precision Rare-Disease Genomics SaMD
Causal Gene+Variant Top-1 Diagnostic Recall (Noisy HPO + VCF)
97.92 %
ClinGen Expert Panel P/LP vs B/LB Concordance AUROC (n=6,011)
0.9814 AUROC
ClinGen Post-2022 Temporal Holdout AUROC (n=4,762, Zero 2022 Overlap Bias)
0.9769 AUROC
Ablation 1: Naive Genotype-Only Variant Top-5 Recall (Without HPO / Trio / Hotspots)
47.08 %
Ablation 2: Preserved 2024–2025 Phenopacket Phenotype-Only Temporal Baseline (172/1,874)
9.18 %
Negative Trio/Singleton Diagnostic Specificity (Posterior < 0.85, n=240)
97.92 %

Interactive Domain Simulation & Cohort Evidence Canvas

PASS
De Novo Trio Diagnostic Hit (RUNX1 Positive) Verdict: PASS · Hydrating interactive scientific canvas...
acmgClass
Pathogenic
acmgPoints
14
causalRankJoint
1
jointPosterior
1
naiveGenoVariantRank
1
phenoOnlyGenomeRank
1

Mathematical / Biophysical Methodology & Upgrade vs APOSTOLUS v1

Combines exact HPO v2022-12-15 ancestor Resnik Information Content across 4,900 disease genes with Tavtigian (2018/2020) Bayesian ACMG/AMP 28-criteria log-likelihood ratios (OP = 350^(pts/8)), Pejaver 2022 ClinGen SVI REVEL/SpliceAI calibrations, trio Mendelian transmission priors, and ISO 14971 Gain-of-Function PVS1 suppression.

Advancement Over Legacy APOSTOLUS v1

Upgrades APOSTOLUS v1 phenotype-only baseline (9.18% Top-5, kept intact as ablation) by integrating patient VCF genotype + ClinGen 8,667 expert-curated variants + Tavtigian Bayesian ACMG point scale.