Anthropic Opens Grant Program for AI Research on Rare Genetic Diseases
Anthropic Opens Grant Program for AI Research on Rare Genetic Diseases
On July 20, Anthropic began accepting applications for research on rare genetic diseases. Selected teams will receive up to $50 thousand in Claude credits for six months. The program has one track for academic researchers and another for biotechnology companies developing therapies.
Data on a rare disease are often scattered across multiple sources. A family may describe the symptoms, a laboratory may identify a gene variant, a registry may contain the patient’s history, and a paper may explain the biological mechanism. The same disease may have several names, while the same symptom may result from different cellular defects. Researchers must therefore translate these records into a shared vocabulary before they can look for connections among them.
For the academic track, Anthropic proposes working with the Monarch Initiative. This project brings together data on diseases, genes, DNA variants, and disease features. Its Mondo ontology maps different names to the same diagnosis. The DisMech database connects symptoms, genetic factors, disease mechanisms, and treatments to specific citations from scientific abstracts.
A participant can ask Claude to identify diseases that share a gene or biological pathway. An expert then reviews the proposed connection and checks the supporting evidence in DisMech. The model proposes a hypothesis, and the expert checks which publications support it. Anthropic also suggests comparing a patient’s features with a diagnosis and investigating a possible mechanism for a genetic variant whose effects on the body remain unclear.
The first track returns its findings to a shared database instead of leaving them in one team’s correspondence. The company plans to publish the results of academic projects on the Monarch website. Future researchers will be able to see the connection, its sources, and how it was verified, rather than starting the search from scratch.
The second track focuses on early therapy development. Anthropic lists several possible tasks: calculating a starting dose using models of how a drug moves through the body, identifying measurable biomarkers in registries and case reports, and preparing sections of a regulatory dossier. In this setting, the model prepares material that a researcher, clinician, or regulator can use to make a decision.
The grant combines access to the model with relevant data, rules for matching those data, and verification of each hypothesis. This sequence is necessary when information about a rare disease is scattered across families, laboratories, registries, and scientific papers.