{
  "id": 425196,
  "title": "Explainable machine learning relates histological to genomic pathology",
  "url": "https://urgent.news/2026/08/09/explainable-machine-learning-relates-histological-to-genomic-pathology",
  "topic": "ai",
  "section": "AI",
  "published": "2026-08-09T00:00:00.000Z",
  "source": {
    "name": "bioRxiv",
    "slug": "biorxiv",
    "url": "https://www.biorxiv.org/content/10.64898/2026.08.03.742582v1?rss=1"
  },
  "original_language": "en",
  "account": "Machine learning techniques are being employed to bridge the gap between histological and genomic pathology in cancer research. By analyzing 597 mouse liver tumors with matched whole-genome sequencing and histopathology, researchers found that deep learning and supervised machine learning models could accurately predict germline and somatic alterations from histological images. These predictions were made at both locus-specific and genome-wide scales, revealing unexpected associations between specific driver mutations and histological features like hepatic steatosis.\n\nHowever, the model's performance varied when applied to tumors from genetically diverse backgrounds, highlighting the importance of considering genetic factors in model development. The study also demonstrated an association between specific genetic alterations and tumor evolution, such as whole-genome duplication. These findings suggest that machine learning can provide interpretable insights into genetic alterations, potentially reducing the need for expensive additional molecular assays. Nonetheless, caution must be exercised when applying these methods to samples with genetic backgrounds beyond those represented in the training data, as the model's performance may be compromised.",
  "summary": "Background & Aims: Haematoxylin and eosin (H&E) staining remains the diagnostic gold standard for solid cancers, including hepatocellular carcinoma, and is increasingly complemented by genomic profiling for precision medicine. Inferring genomic alterations directly from H&E images could streamline testing, but heterogeneity and biases in human training data limit interpretation of…",
  "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."
}