The 8 PASS-quality Mutect2 calls from a real, ENA-extracted tumour/normal run, cross-checked against two independent sources: the NYGC published truth set (via som.py) and Cellosaurus's curated COLO829 genotype (CVCL_1137).
| Coordinate (hg38) | Change | Truth-set (som.py) | Independent cross-check | Status |
|---|---|---|---|---|
| chr7:140,753,336 | A>T | TP | Cellosaurus + OncoKB + GeneBe — exact match to documented BRAF p.Val600Glu | Driver — BRAF V600E |
| chr9:21,971,154 | CCG>C | TP | Cellosaurus — exact match to documented CDKN2A p.Ala68Glyfs*51 | Driver — CDKN2A frameshift |
| chr5:1,295,113 | GG>AA | TP | Cellosaurus — reverse-complement match to documented TERT promoter CC>TT (UV signature) | Driver — TERT promoter |
| chr7:140,800,999 | G>A | TP | Not individually named in Cellosaurus's summary; falls inside BRAF gene body (MODIFIER impact) | Truth-set confirmed |
| chr7:140,803,447 | G>T | TP | Not individually named in Cellosaurus's summary; falls inside BRAF gene body (MODIFIER impact) | Truth-set confirmed |
| chr17:7,668,302 / 7,668,311 | phased | TP | Not corroborated by Cellosaurus; falls inside TP53 (MODIFIER impact — no TP53 mutation documented for this line) | Truth-set confirmed |
| chr17:7,668,450 | — | TP | Not corroborated by Cellosaurus; falls inside TP53 (MODIFIER impact) | Truth-set confirmed |
Recall against the truth set reads near-zero by design: the truth VCF is whole-genome (41,427 SNVs, 984 indels) while this run's query is deliberately scoped to ~30kb across 8 melanoma driver genes. Precision — the number that means something at this scope — is 1.0 across every variant class. A real, disclosed gap: COLO829's documented PTEN c.493_634del142 (142bp deletion) was not recovered — outside what an SNV/indel-tuned caller like Mutect2 reliably assembles.