A San Francisco genomics firm has secured up to $10M from ARPA-H to expand an AI platform built to diagnose rare genetic disease. The company, Probably Genetic, disclosed the award on August 31. The money flows through RAPID, the federal agency’s rare disease diagnostics program.
The program attacks the long diagnostic odyssey facing patients with rare conditions. Roughly 10,000 diseases qualify, they touch more than 350 million people worldwide, and the average search for answers runs six years, ARPA-H estimates. Since December 2024 the agency has funded AI detection models and assembled a large curated dataset of longitudinal patient data for training and benchmarking.
Probably Genetic’s job is gathering data at scale. It will enroll people with confirmed diagnoses and pool clinical records, patient-reported information and biological data, including DNA, into one de-identified dataset built on informed consent. AI models trained on it should help flag undiagnosed patients and feed RAPID’s national data ecosystem.
The company’s direct-to-patient model already lets people order genetic testing and counseling without a doctor’s referral. This contract pushes that model toward the research side, turning patient-driven data into fuel for the next generation of rare disease diagnostics.
