{
  "id": 9302968,
  "title": "AI scans 4.6 million compounds in hours to predict hydrogen positions in drug-like molecules",
  "url": "https://urgent.news/2026/09/23/ai-scans-4-6-million-compounds-in-hours-to-predict-hydrogen-positions",
  "topic": "ai",
  "section": "AI",
  "published": "2026-09-23T09:00:02.000Z",
  "source": {
    "name": "Phys.org",
    "slug": "phys-org",
    "url": "https://phys.org/news/2026-09-ai-scans-million-compounds-hours.html"
  },
  "original_language": "en",
  "account": "New York University scientists have developed an artificial intelligence (AI) model capable of rapidly predicting the most stable position of hydrogen atoms in drug-like molecules. The research, published in the journal Chemical Science, addresses a long-standing challenge in molecular design and drug discovery: determining the correct tautomer form of molecules that can exist in multiple related structures.\n\nTautomers occur when a hydrogen atom shifts between positions, altering the molecule's bonding pattern. These seemingly small changes can significantly impact how a molecule interacts with proteins, making accurate tautomer assignment crucial for molecular modeling and structure-based drug discovery. However, determining the correct tautomer is difficult due to limited experimental data and the computational expense of other methods.\n\nTo tackle this issue, the researchers mined the Cambridge Structural Database, a repository containing high-resolution small-molecule X-ray crystal structures with known hydrogen positions. They created a dataset of over 1.1 million tautomeric states and trained a graph neural network, a type of AI that can identify patterns in connected data points, to predict stable tautomers directly from 2D molecular structures.\n\nWhen applied to 5,075 protein-bound ligands from the Protein Data Bank, the model identified 126 cases where the previously assigned tautomers were likely incorrect. The AI model consistently reassigned alternative stable tautomers, improving hydrogen-bonding patterns and providing more chemically reasonable interactions. While the experimentally determined protein structures are likely accurate, the AI suggests that the chemical representation of these tautomers may need revision.\n\nFor example, in one case, the AI model predicted a tautomer with a differently positioned hydrogen atom than the original assignment. This change led to the ligand forming additional hydrogen bonds with nearby protein residues, demonstrating how a small change can significantly alter the molecule's interactions with its protein environment. Accurately identifying tautomers is essential for drug discovery, as it can help predict how a drug candidate fits into a protein binding site and improve computer simulations of molecular dynamics. The researchers released their method as an open-source tool called Tautomer-Predictor, which can rapidly analyze large molecular libraries.",
  "summary": "New York University researchers have trained an AI model to learn chemical patterns associated with stability in drug-like molecules and accurately predict where their hydrogen atoms should be positioned.",
  "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."
}