{
  "id": 18169945,
  "title": "Neural RNA splicing models rely on reading frame information unavailable to the splicing machinery",
  "url": "https://urgent.news/2026/11/04/neural-rna-splicing-models-rely-on-reading-frame-information",
  "topic": "ai",
  "section": "AI",
  "published": "2026-11-04T00:00:00.000Z",
  "source": {
    "name": "bioRxiv",
    "slug": "biorxiv",
    "url": "https://www.biorxiv.org/content/10.64898/2026.11.03.763976v1?rss=1"
  },
  "original_language": "en",
  "account": "Neural networks that predict RNA splice sites based on genomic sequences may not be accurately replicating the natural splicing process, according to recent research. The study focused on mutations that do not significantly impact splicing but are correlated with it in the human genome, particularly those that affect the reading frame alignment. The researchers discovered that neural tools commonly used for this task are significantly more affected by mutations that alter the reading frame, resulting in in-frame stop codons, compared to those that maintain the reading frame. This finding suggests that these models are missing many \"poison exons,\" which are clinically relevant categories where the reading frame is disrupted. The dependence on reading frame was found to be much weaker in \"grey-box\" models that incorporate splicing-specific structure. The study's findings indicate that widely-used black-box models of splicing are relying on information that is not accessible to the RNA splicing machinery itself. The code used in this research is available for further examination at the provided GitHub link.",
  "summary": "Deep neural networks can accurately predict RNA splice sites from genomic sequence, but whether they do so by simulating the natural splicing process is unclear. We evaluate this question by examining mutations that should not have a strong effect on splicing, but are correlated with it in the human genome. We focus specifically on mutations that affect frame alignment because this is known to be…",
  "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."
}