{
  "id": 17331922,
  "title": "Agentic cryo-EM structure determination with CryoPilot",
  "url": "https://urgent.news/2026/10/30/agentic-cryo-em-structure-determination-with-cryopilot",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-30T00:00:00.000Z",
  "source": {
    "name": "bioRxiv",
    "slug": "biorxiv",
    "url": "https://www.biorxiv.org/content/10.64898/2026.10.29.762908v1?rss=1"
  },
  "original_language": "en",
  "account": "CryoEM's single-particle technique relies on computational tools, but human experts must still review intermediate results, choose between processing options, and determine when changes are needed. Researchers have now developed CryoPilot, a system utilizing two language model agents to guide cryo-EM data from raw micrographs to validated atomic models. Unlike a predetermined pipeline, CryoPilot's reconstruction agent assesses intermediate images and diagnostic metrics, selects and configures CryoSPARC jobs, tests alternative approaches, compares results, and retains, revises, or reverts decisions as necessary. Both agents operate within a persistent programming environment and session database, preserving the entire processing history and enabling multi-day workflows with minimal human intervention.\n\nAcross nine public EMPIAR datasets, CryoPilot achieved reconstructions with resolutions ranging from 2.61 to 3.51 Å. When refining models from seven maps, seven reconstructions aligned with reference structures within 0.4 to 4.9 Å Carra. For a particularly challenging case involving an alpha latrocrustotoxin, a single modification to the particle picking strategy recommended by CryoPilot improved the reconstruction from 5.9 to 4.15 Å. By establishing an adaptive and auditable framework for agent-driven cryo-EM structure determination, CryoPilot streamlines the process while providing extensive traceability and the ability for the system to learn and adapt over time.",
  "summary": "Single-particle cryo-electron microscopy (cryo EM) relies on mature computational tools, but experts must still inspect intermediate results, choose among alternative processing routes, and decide when earlier choices should be revised. Here we present CryoPilot, a system of two language model agents that carries cryo-EM data from raw micrographs through three-dimensional reconstruction to…",
  "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."
}