Probably Genetic Wins Up to $10M ARPA-H Contract

Probably Genetic has been awarded up to $10 million by the Advanced Research Projects Agency for Health (ARPA-H) to expand its AI platform for diagnosing rare genetic diseases, the San Francisco company announced on August 31, 2026. The award was made under ARPA-H’s Rare Disease AI/ML for Precision Integrated Diagnostics (RAPID) program, a federal effort to shorten the years-long search many patients face before receiving an accurate diagnosis.

ARPA-H is an agency within the U.S. Department of Health and Human Services that funds high-risk, high-reward biomedical research. Its RAPID program, launched in December 2024, aims to develop AI-based detection models for rare and ultra-rare diseases and to assemble a large curated dataset of longitudinal patient data for training and benchmarking diagnostic algorithms. According to ARPA-H’s program page, more than 10,000 rare conditions collectively affect over 350 million people worldwide, and the search for a diagnosis lasts six years on average.

What the Award Funds

Under the contract, Probably Genetic will scale its direct-to-patient data platform across hundreds of rare diseases. The company said it will recruit individuals who already have established diagnoses and meet RAPID’s data and participation criteria, aggregating clinical records, patient-reported information, and biological data including DNA into a single dataset. The company said the data will be de-identified and used only for purposes to which patients gave explicit informed consent.

Probably Genetic said the resulting resource will be used to train AI models that identify undiagnosed patients and will serve as a component of RAPID’s broader national-scale data ecosystem for developing and benchmarking diagnostic algorithms. Beyond diagnosis, the company said the dataset will support multi-omic phenotype models intended to link real-world evidence to disease biology, giving drug developers information to identify targets, stratify patients, and design clinical trials.

“Rare disease diagnosis remains one of medicine’s most difficult and under-addressed challenges,” said Scott Gorman, RAPID Program Manager at ARPA-H, in the company’s announcement. He said the program pairs novel AI approaches with multimodal data to enable cross-disease detection at greater speed and scale while generating insights intended to accelerate treatment development.

The Data Problem in Rare Disease Diagnosis

The company said half of rare disease patients remain undiagnosed, and that AI approaches to date have been constrained by fragmented, low-quality data. It pointed to an overreliance on electronic health records that often lack details such as symptom onset, severity, progression, and morphological features.

Probably Genetic’s platform converts self-reported symptoms, prior clinical diagnoses, electronic health record data, and patient-submitted material such as photos and videos into structured deep phenotypic data. The company also provides at-home genetic testing, which it said reduces reliance on repeated specialist visits.

The company said it has collected data from more than 120,000 patients to date and partnered with more than 50 patient advocacy groups and more than 15 biopharmaceutical companies. Lange said that if the effort succeeds, the company will assemble what he described as the largest AI-ready genetic disease dataset to date, a characterization the company presented as its own objective.

Program Context

RAPID is one of several ARPA-H programs directed at rare disease. According to the agency, the program’s strategy is to build provider-facing diagnostic tools that integrate into existing clinical workflows, alongside cost-effective direct-to-patient systems that can be deployed remotely to help individuals detect rare diseases at home or in non-clinical settings and route them toward appropriate medical support. The agency said that if the program succeeds, it will expand access to rare disease expertise and help patients and providers reach an accurate diagnosis faster than is possible today.

Probably Genetic was founded in 2018 and describes itself as an AI platform for the research, diagnosis, and treatment of genetic diseases. The company said its mission is to diagnose more than 200 million patients living with genetic disease and to support the discovery and development of treatments using AI. The company stated that the research is funded in part by ARPA-H and noted that the views expressed in its materials are those of the authors and do not represent official U.S. government policy.

Under the RAPID award, Probably Genetic said it will deploy its patient data submission portal across hundreds of rare diseases, recruiting diagnosed patients to build a dataset it described as large, diverse, and deeply characterized. The company said this resource will feed the development, training, and benchmarking of diagnostic algorithms designed to shorten — and ultimately end — the rare disease diagnostic odyssey.