Sage Bionetworks to build rare disease data platform for ARPA-H

Sage Bionetworks will build the Rare Disease Data Commons for ARPA-H’s new RAPID program, with funding of up to $28 million over 4.5 years. The platform is meant to unify fragmented rare disease data, create benchmark datasets and support AI tools aimed at faster diagnosis.

Why it matters: - Rare disease diagnosis often takes about six years, and some families wait decades for answers. - More than 10,000 rare diseases affect roughly 30 million Americans. - Rare diseases cost the U.S. health system an estimated $1 trillion a year. - Only about 5% of rare diseases have an approved treatment. - The new data infrastructure is designed to speed diagnosis, reduce costs and improve evaluation of AI tools.

What happened: - Sage Bionetworks is partnering with the Advanced Research Projects Agency for Health, or ARPA-H, to build the Rare Disease Data Commons. - The Rare Disease Data Commons will serve as the central data repository and benchmarking platform for ARPA-H’s Rare Disease AI/ML for Precision Integrated Diagnostics, or RAPID, program. - ARPA-H is awarding Sage Bionetworks up to $28 million over 4.5 years. - The work is based in Seattle. - Sage Bionetworks is building the platform with Cirro Bio, Netrias, the Gyori lab at Northeastern University, the Alsentzer lab at Stanford University and the Undiagnosed Diseases Network Foundation. - Organizations interested in contributing data or testing methods can contact the Sage Bionetworks team at rapid@sagebionetworks.org.

The details: - RAPID has four technical areas. - The first technical area focuses on assembling a large-scale longitudinal clinical dataset from electronic health records. - The second technical area will collect patient-reported data and other modalities, including genomics, imaging, video, voice and wearable data. - Sage Bionetworks leads the third technical area, which will harmonize incoming data, govern access and benchmark diagnostic AI built on top of it. - The fourth technical area will validate tools in clinical settings. - Patient experience partners work across all four areas. - The Rare Disease Data Commons will contain five connected modules. - The platform will bring in data from the other RAPID teams and outside contributors. - The system will standardize data enough to compare across sources, building on tools from the ARPA-H Biomedical Data Fabric Toolbox. - The platform will set rules for who can use the data and how. - When data cannot be centralized, the Rare Disease Data Commons will support federated integration and interoperable connections with other trusted research environments. - The Rare Disease Benchmark Dataset will be versioned and broadly accessible. - The benchmark dataset is intended to give AI developers a common basis for evaluation. - The benchmark will use de-identification, governed access and data-use controls to protect patient privacy. - Sage Bionetworks also develops and operates Synapse, an NIH-listed generalist repository that manages more than 4.5 petabytes of biomedical research data. - Sage says the Rare Disease Data Commons builds on 15 years of experience developing governed data commons for sensitive biomedical data.

Between the lines: - The core problem is not a lack of AI methods, but a lack of usable data. - Rare disease data is scattered across institutions, coded differently and often locked behind separate governance terms. - The platform is meant to connect data across diseases so researchers can look for shared and latent patterns. - A common benchmark could make it easier to compare AI methods and determine which tools are worth clinical follow-up. - The effort reflects a push to build open infrastructure without weakening privacy protections. - Robert Allaway, Sage Bionetworks’ director of rare disease, said the goal is to connect separate data “islands” into an “archipelago.” - Milen Nikolov, Sage Bionetworks’ director of product innovation, said the modules are meant to turn disjoint multi-modal datasets into information usable by patients, researchers and clinicians. - Jineta Banerjee, Sage Bionetworks’ associate director of advanced data analytics, said the project aims to learn across rare diseases to uncover patterns that could improve diagnosis or identify therapeutic opportunities. - Luca Foschini, Sage Bionetworks’ president and CEO, said building in the open while protecting privacy is necessary to create trust and share learnings.

What's next: - Sage Bionetworks will continue building the Rare Disease Data Commons as RAPID moves through its four technical areas. - The benchmark dataset is expected to become a public-facing resource for evaluating diagnostic AI. - The platform will support future contributions of data and method-testing from outside partners. - The system will be used to validate tools in clinical settings as the program advances.

The bottom line: - ARPA-H is backing a shared data layer to tackle one of rare disease care’s biggest bottlenecks: too much fragmented data, and too little standardization to turn it into faster diagnoses.