{
  "id": 3043861,
  "title": "Gene identity, not variant effect, dominates ClinVar benchmarks of missense pathogenicity predictors",
  "url": "https://urgent.news/2026/08/24/gene-identity-not-variant-effect-dominates-clinvar-benchmarks-of",
  "topic": "science",
  "section": "Science",
  "published": "2026-08-24T00:00:00.000Z",
  "source": {
    "name": "bioRxiv",
    "slug": "biorxiv",
    "url": "https://www.biorxiv.org/content/10.64898/2026.08.21.746180v1?rss=1"
  },
  "original_language": "en",
  "account": "A study examines the dominant factor in ClinVar benchmarks for missense pathogenicity predictors, finding that gene identity, rather than variant effect, plays a crucial role. The researchers analyzed 197,904 ClinVar missense variants validated against UniProt canonical sequences and found that a null model, which scores each variant based on the pathogenic fraction of its gene, achieves an area under the receiver operating characteristic curve (AUROC) of 0.921.\n\nWhen compared to four current predictors, the difference in AUROC ranges from 0.036 to 0.044, indicating that gene identity alone accounts for a significant portion of the predictors' performance. However, the inflation in performance is not uniform across predictors. Within-gene evaluation reveals that the ranking of predictors changes, with AlphaMissense rising to first place and gMVP falling to third place (p).",
  "summary": "Missense pathogenicity predictors are routinely benchmarked against ClinVar, whose labels are strongly structured by gene: genes under diagnostic scrutiny accumulate pathogenic submissions while incidentally sequenced genes accumulate benign ones. We asked how much of a benchmark score this structure alone can produce. On 197,904 ClinVar missense variants validated against UniProt canonical…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}