By systematically testing every possible single mutation, they were able to build a detailed functional map of how genetic variations affect protein expression and function. Building on this work, they expanded their experiments to a crucial liver transporter, responsible for metabolizing a range of pharmaceuticals. The research team studied more than 14,000 changes in DNA—like missing or extra pieces—to build a powerful dataset that connects how genes behave with how proteins actually work in the body.
AI-Powered Discovery: The CPT Model On the computational front, the team developed CPT (Cross-Protein Transfer), a variant effect predictor that integrates unsupervised protein sequence models, structural insights, and evolutionary data. CPT has already set a new benchmark in the field—recent independent testing by Joseph Marsh’s lab at the University of Edinburgh ranked it as the top-performing tool for deep mutational scanning data, surpassing even DeepMind’s AlphaMissense. “The predictive models we’ve developed provide a much more comprehensive understanding of how genetic variants impact solute carrier transporters,” Song said. “By leveraging AI and computational modeling, we can move toward a future where treatments are designed specifically for each patient’s genetic blueprint, rather than a one-size-fits-all approach.” Stock image: DNA helix
THE SUPPORT FROM THE NOYCE INITIATIVE HAS BEEN INSTRUMENTAL IN EXPANDING OUR RESEARCH SCOPE AND FOSTERING INTERDISCIPLINARY PARTNERSHIPS.
The Future of Predictive Modeling in Medicine By advancing machine learning models capable of predicting disease-related mutations, the team is laying the foundation for breakthroughs in precision medicine, drug development, and diagnostics. While their current focus is on solute carrier transporters, the methods they’ve pioneered can be adapted to study other essential proteins with implications for a wide range of diseases. As the research moves forward, the team is actively working to refine their computational models and expand their functional assays to additional transporters. With the continued support of the UC Noyce Initiative, their work promises to redefine how genetic mutations are understood and treated, ultimately leading to better health outcomes for patients worldwide. ◆ A Collaboration That’s Transforming Computational Health This ambitious research project would not have been possible without the UC Noyce Initiative, which fostered collaboration between computational scientists and experimental biologists across the UC campuses, according to the team. The funding allowed the team to: Expand functional assays beyond their initial studies on organic cation transporter member 1 to other solute carrier transporters. Train and mentor three Ph.D. students and one postdoc at UC Berkeley, along with an assistant researcher at UCSF. Launch a new collaboration with Willow Coyote-Maestas’ lab at UCSF, broadening their research into G protein-coupled receptors, another class of clinically significant proteins. “The support from UC Noyce has been instrumental in expanding our research scope and fostering interdisciplinary partnerships,” Song said. “It has enabled us to bridge the gap between computational modeling and biological experimentation, accelerating discoveries that will directly impact human health.”
62 Impact Report 2023 - 24 | UC NI
Made with FlippingBook Annual report maker