All of Us Wearable Data Pre-Processing for Disease Risk Prediction Modeling | Office Hours 7.31.26

In this video, our team is joined by the Machine Learning for Health (ML4H) team from the Broad Institute. The team will share how they use All of Us wearable, electronic health record, and genomic data to study disease risk and progression, including a reproducible pipeline for data ingestion, quality control, phenome-wide prediction, and longitudinal modeling with transformer-based methods. They will also highlight key findings showing how wearable features improve disease prediction beyond demographics and how combining data types can strengthen clinical insight.

Speakers include Sam Freesun Friedman, a senior group leader and principal machine learning scientist; Tal Shnitzer, a senior machine learning scientist; and Valentina D’Souza, a machine learning engineer, all at the Broad Institute. Together, they bring expertise in developing clinically grounded AI models and translating complex biomedical data into actionable insights. The team will also provide updates on genomic integration, benchmark development, and reproducible notebooks for the All of Us community.

recorded July 31, 2026