RaDiCal benchmark data — cohort manifests, benchmark results, head-to-head experiment, and pre-registration document (v2)

Benchmark data supporting the RaDiCal preprint: RaDiCal: a phenotype-robust multi-engine platform for rare-disease, mitochondrial, and secondary-findings variant interpretation, validated on public benchmark cohorts (Vernon Crock, Virtual Cognition Pty Ltd).This dataset (v2) contains 11 files: COHORT.json v2.0 — 200-case GA4GH Phenopacket Store v0.1.26 benchmark cohort manifest (seed=42) RESULTS_reranker_am_final.json — Baseline 2: ideal-mode 200-case benchmark results (192/200 = 96.0%) RESULTS_degraded_50pct.json — Baseline 6: degraded 50% HPO dropout 200-case results (192/200 = 96.0%, Δ=0.0 pp) RESULTS_comparatorA_vestito_hth.json — Head-to-head Comparator A: Cooperstein-Exomiser configuration (192/200 = 96.0%) RESULTS_comparatorB_rerun.json — Head-to-head Comparator B: RaDiCal default (192/200 = 96.0%) RESULTS_comparatorC_ablation.json — Head-to-head Comparator C: Cooperstein config + AlphaMissense reranker ablation (192/200 = 96.0%) vestito-headtohead-mcnemar.json — McNemar analysis output (mid-p p=1.000, Δ=0.00 pp, pre-registered) cooperstein-headtohead-PREREG.md — Pre-registration document (git commit 7983951ec788a1100eaa5c5e7c4d2677b484bdcf) MITO_COHORT_v2.json — Mitochondrial tuning cohort manifest (seed=42, n=50: 25 spiked / 25 controls) MITO_COHORT_v3.json — Mitochondrial held-out validation cohort manifest (seed=1337, n=50) SF_COHORT.json — Secondary-findings engine benchmark cohort manifestAll files CC BY 4.0. Platform git commits: f1c7a81fe3ae9def1e9b5f25c250a17e7aaf934f (Baseline 6), 25b6426f68c5f831e7545b808284ae2702892b87 (head-to-head experiment).