Anthropic AI for Science Rare Disease Research Grants

Anthropic has introduced a focused call for applications for rare disease research grants as part of its broader AI for Science program. The initiative provides accepted applicants with up to $50,000 in Claude credits over six months to build a community of researchers using AI to improve the understanding and treatment of rare genetic diseases.

Addressing Challenges in Rare Disease Research

Rare genetic diseases present significant hurdles for traditional research due to fragmented data and small patient populations. With an estimated 400 million people living with more than 7,000 rare diseases, the scale of the problem is vast, yet the knowledge for any single condition is often limited.

Key challenges identified by Anthropic include:

Data Fragmentation: Small, scattered populations make it difficult to build patient registries and identify therapeutic targets.
Isolation of Study: Unique genetic variations are often studied in isolation, preventing the identification of shared mechanisms across different diseases.
Guidelines and Documentation: The time required to move drug candidates into patient trials is often extended by the manual assembly of thousands of pages of regulatory documentation.

Anthropic posits that AI can mitigate these issues by modeling rare genetic diseases, detecting patterns across disparate conditions, synthesizing findings from large literature corpora, and creating shared terminology to maximize the utility of limited datasets.

Track One: Scaling Basic Science Partnerships

Track one focuses on collaboration between clinical researchers, patient organizations, and data scientists to accelerate the discovery of mechanisms underlying rare diseases.

Integration with the Monarch Initiative

Anthropic is partnering with the Monarch Initiative, an international consortium that provides critical resources for rare disease research, including:

Mondo Disease Ontology: A computational framework that reconciles disease definitions from multiple sources like OMIM, Orphanet, and ICD.
Monarch Knowledge Graph: A system that integrates genotype-phenotype data across species.
DisMech: A new agent-friendly mechanistic disease classification library. Claude is used within DisMech to analyze case reports, variant databases, and registry schemas to identify mechanistic similarities between diseases at scale.

Outputs from track one projects will be made publicly available at Monarchinitiative.org.

Track Two: Scaling Biotech Partnerships

Track two supports early-stage biotechs and biotechnologists aiming to compress the timeline from genetic diagnosis to available treatment, which currently takes one to two years.

Accelerating Drug Development

Anthropic suggests that Claude can be used to:

Streamline Documentation: Draft and review regulatory dossiers to reduce the time spent on manual assembly.
Optimize Therapeutic Strategy: Analyze target druggability across various modalities (e.g., antibodies, small molecules, genetic medicines).
Identify Shared Mechanisms: Find commonalities across genetic therapies to potentially allow for "basket trials," reducing the need for separate Investigational New Drug (IND) filings for each patient.

Existing AI for Science Partners

Several organizations are already utilizing Claude for rare disease applications:

Every Cure: Identifying drug repurposing opportunities across millions of candidates.
Centre for Population Genomics: Drafting variant classifications for expert review to remove diagnostic bottlenecks.
Violet Research Institute: Navigating FDA guidelines, running bioinformatics pipelines, and drafting regulatory filings for ultra-rare genetic diseases.

Application Details and Limitations

Applications are accepted through August 2, 2026, at 11:59 PM PST. Accepted applicants can use credits for Claude Opus or other approved biological models, with potential exemptions for projects that trigger bio classifiers.

Program Limitations

Anthropic acknowledges that AI cannot solve all problems in this field, specifically:

Data Scarcity: AI cannot function where data is too paltry or poorly organized for agents to reach.
Infrastructure: AI cannot address systemic issues such as insurance authorization or access to diagnostic facilities.

Example Project Scopes

Track One Examples:

Ranking mechanistic links between rare diseases sharing a gene or pathway using DisMech.
Curating patient organization data to improve natural history studies.
Building evaluations to measure model performance on phenotype-to-disease matching and mechanism prediction.

Track Two Examples:

Synthesizing PK/PD modeling and allometric scaling to justify starting doses from sparse data.
Mining case reports to identify biomarkers and functional endpoints for N-of-1 programs.
Drafting and cross-checking regulatory documentation (CMC modules, IND sections) to compress assembly time.