{
  "id": 2234806,
  "title": "CryoForge: A Self-Correcting Agent for Cryo-EM Model Building That Learns When to Act and When to Stop",
  "url": "https://urgent.news/2026/08/20/cryoforge-a-self-correcting-agent-for-cryo-em-model-building-that",
  "topic": "ai",
  "section": "AI",
  "published": "2026-08-20T00:00:00.000Z",
  "source": {
    "name": "bioRxiv",
    "slug": "biorxiv",
    "url": "https://www.biorxiv.org/content/10.64898/2026.08.15.745007v1?rss=1"
  },
  "original_language": "en",
  "account": "CryoForge is a novel agent designed for cryo-electron microscopy (cryo-EM) model building that learns when to make corrections and when to stop. Current automated builders generate distinct residual error profiles that necessitate expert evaluation. The main challenge lies in determining which local interpretations are justified by experimental data and should be kept, adjusted, or discarded.\n\nCryoForge addresses this challenge by acting as a post-builder agent that separates the decision-making process for repair proposals from the acceptance of these repairs. It employs rule-based and learning-based components to identify regions that require repair and prioritize appropriate actions. An independent evidence gate assesses each edit based on map and half-map support, stereochemistry, connectivity, and local structural context. Edits that are supported are retained, while those that are unsupported or conflicting are rejected, rolled back, stopped, or escalated for expert review.\n\nIn a resolution-stratified benchmark, CryoForge demonstrated significant improvements. Out of 26,153 released trajectories, 84.9% yielded standard validated enhancements, and 3.7% experienced low-confidence partial improvements. Importantly, no quality-degrading edits were retained in the final promoted models. Compared to a rule-only control, CryoForge's learned prioritization reduced non-improving candidates and harmful actions while maintaining global structural stability. External evaluations using alternative initializers, team submissions involving both automated and manual corrections, and three recently released complex assemblies further confirmed CryoForge's adaptability to different residual error patterns. The agent performs bounded, evidence-supported corrections without the risk of uncontrolled remodeling. Ultimately, CryoForge offers a builder-independent, scalable, and auditable correction layer between automated model generation and expert structural interpretation.",
  "summary": "Automated atomic model building has accelerated cryo-EM structure determination, but different builders leave distinct residual error profiles requiring expert inspection. The post-building challenge is to decide which local interpretations are sufficiently supported by experimental evidence to be retained, corrected or rejected. Here we introduce CryoForge, an evidence-gated post-builder agent…",
  "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."
}