A call for projects targets rare diseases with API funding

Anthropic is awarding $50,000 in API credits to rare disease projects. A lever to unify data and draft clinical records.

Anthropic is opening a call for applications dedicated to rare genetic diseases, as part of its AI for Science program. Selected projects will receive up to fifty thousand dollars in credits over six months, split into two tracks: one for foundational research, and the other for early-stage biotechs working on clinical development. Applications are open until August 2.

The scope of the program addresses a structural challenge in the field. Approximately four hundred million people live with one of the seven thousand identified rare diseases, but these dispersed populations complicate the creation of registries, the identification of therapeutic targets, and the design of clinical trials. Because each disease is studied through its own specific characteristics, such as genetic variations or symptom combinations, shared mechanisms across pathologies go unnoticed.

The first track builds on the Monarch Initiative, an international consortium that produces interoperability resources like the Mondo ontology, which reconciles disease definitions scattered across OMIM, Orphanet, ICD, and other sources. Its contributors are assembling a mechanistic classification library called DisMech, designed to be agent-actionable. The work resulting from this track will be made publicly available on the Monarch Initiative website.

The second track aims to reduce development timelines, which currently stand at one to two years between a confirmed genetic diagnosis and an accessible treatment—a duration partly consumed by backlogs in certified manufacturing, sequential safety studies, and the manual assembly of thousands of pages of regulatory documentation. Suggested avenues of research include the drafting and reviewing of regulatory filings, the selection of therapeutic strategies, and the search for common mechanisms that would allow for a basket trial rather than a patient-by-patient filing.

Anthropic specifies the limitations of this approach: models cannot help where data is missing or remains too poorly organized, and they do not address healthcare access barriers such as insurance approvals or the availability of diagnostic infrastructure.