The federal government is launching a major initiative to use artificial intelligence and machine learning to diagnose rare diseases faster, potentially cutting through the diagnostic delays that currently leave patients waiting years for answers. The Advanced Research Projects Agency for Health (ARPA-H), part of the U.S. Department of Health and Human Services, announced contract awards through the Rare Disease AI/ML for Precision Integrated Diagnostics (RAPID) program, which will invest up to $98.5 million over 4.5 years to develop AI-powered diagnostic tools and build foundational data resources for rare disease research.

Why Is Rare Disease Diagnosis Such a Massive Problem?

More than 350 million people worldwide live with one of over 10,000 rare diseases, yet the path to diagnosis remains painfully slow. Patients with rare diseases face an average diagnostic odyssey lasting six years, though some wait decades for a definitive answer. This delay isn't just frustrating; it has serious medical consequences. Prolonged diagnostic uncertainty often leads to inappropriate care, irreversible disease progression, and rising medical costs. The bottleneck also slows treatment development because researchers struggle to recruit enough patients for clinical trials when diseases remain undiagnosed. Currently, only about 5% of rare diseases have an approved therapy, underscoring how urgently the field needs better diagnostic and research tools.

The challenge is compounded by the limitations of current technology. While genomic sequencing has improved diagnostic rates for many rare diseases, many genetic test results remain inconclusive without additional clinical context. Other rare diseases lack a known genetic cause entirely, leaving patients without any definitive diagnosis despite extensive testing.

How Will AI Change Rare Disease Diagnosis?

The RAPID program aims to build what will become the largest AI-ready, real-world data resource of its kind for rare diseases, linking longitudinal clinical records with genomic data, patient-reported information, and other emerging health data sources. This unified dataset will serve as the foundation for developing AI-driven diagnostic tools that can drastically reduce time to diagnosis and accelerate clinical trials through precise patient cohort identification.

Four leading teams have been selected to build different components of this ecosystem:

• University of North Carolina: Building the largest real-world rare disease dataset by integrating clinical, genomic, and patient-reported data across thousands of rare diseases, creating a privacy-preserving foundation for research and innovation.
• Sage Bionetworks: Developing a secure and interoperable platform that harmonizes multimodal rare disease data and enables development, evaluation, and benchmarking of new AI approaches for diagnosis and discovery.
• FDNA: Creating tools to collect longitudinal, multimodal health data including photos, videos, voice recordings, and patient-reported concerns directly from patients at national scale, then deploying AI tools in real-world clinical and direct-to-patient settings.
• Probably Genetic: Developing patient-centered tools to collect multimodal data and generate high-fidelity synthetic datasets to support rare disease research, while creating patient-facing AI tools for earlier disease identification.

The program will also launch Rare Challenges, an open innovation platform that uses rigorous competitions to benchmark and accelerate emerging AI approaches across diagnosis and mechanistic understanding, enabling researchers beyond the core performer teams to contribute solutions.

Steps to Accelerate Rare Disease Diagnosis Through AI

• Standardize and unify data: Remove barriers preventing rare disease data and technologies from being shared across institutions through standardized formats, privacy-preserving data sharing protocols, and improved consent management.
• Develop AI-ready datasets: Build large, representative datasets that AI systems can learn from, addressing the current shortage of robust data that has limited progress in AI-powered diagnostic tool development.
• Create patient-centered tools: Design AI systems that collect multimodal health information directly from patients, including photos, videos, wearable data, and functional assessments, enabling earlier identification of diseases with complex or variable presentations.
• Benchmark and validate solutions: Establish transparent performance assessment standards and shared infrastructure through competitive challenges to accelerate innovation across the rare disease ecosystem.

"For millions of patients and families, living with a rare disease still means years without an accurate diagnosis and too few paths to effective treatment. RAPID aims to change that trajectory by building the capabilities needed to improve diagnosis and support treatment development across thousands of rare and ultra-rare conditions," said Scott Gorman, RAPID Program Manager.

Scott Gorman, RAPID Program Manager at ARPA-H

The initiative reflects a broader recognition that AI can help address the data scarcity problem that has long hampered rare disease research. Current AI-powered diagnostic tools could identify rare diseases earlier and broaden access to diagnostic expertise, but progress has been limited by the shortage of robust, representative datasets needed to develop and validate these tools.

Major technology companies are backing the effort. OpenAI, Anthropic, Amazon Web Services, and Google have pledged in-kind support for RAPID, including computing credits for large language models (LLMs), which are AI systems trained on vast amounts of text data to understand and generate human language. Additional support comes from the Lawrence Berkeley National Laboratory, the National Institutes of Health's All of Us Center for Linkage and Acquisition of Data, and the Monarch Initiative.

Patient advocacy groups are also central to the program's success. Leading organizations including Global Genes and the National Organization for Rare Disorders (NORD) will collaborate closely with technical performers to ensure that the solutions developed reflect the actual needs of the rare disease community.

The RAPID program represents a significant shift in how the U.S. government is approaching rare disease research. Rather than funding isolated studies, ARPA-H is building shared infrastructure and datasets that multiple teams can leverage, creating a more collaborative ecosystem. This approach could accelerate not just diagnosis, but also clinical trial design, endpoint selection, and therapeutic development across the rare disease landscape.