CAREER: Decoding gene regulatory mechanisms with AI and population-scale multi-omics

Modern biology has entered an era in which the central challenge is not the lack of data, but how to organize large and complex datasets into knowledge that scientists can understand and test. Artificial Intelligence has greatly improved the ability to find patterns in biological data, but many models still struggle to explain the biological mechanisms behind those patterns. This project addresses that challenge by developing new ways to use Artificial Intelligence to understand gene regulation, the process by which cells turn genes on and off in response to internal and external signals. The key idea is that protein structure can provide the missing biological context. Mutations that occur near one another in the folded shape of a protein may affect the same functional surface, interaction site, or regulatory region, leading to shared effects on gene activity. By using protein shape to organize genetic and molecular data, this project aims to make Artificial Intelligence models more interpretable, more biologically grounded, and more useful for discovery. The project advances current National Science Foundation priorities in Artificial Intelligence and biotechnology by developing explainable computational tools for biological discovery and using genome-editing experiments to test model predictions. The project also includes educational activities that expand access to Artificial Intelligence and biological data literacy training. Through interactive tutorials, workshops, and research experiences using real biological datasets, students will build skills in coding, data analysis, and interdisciplinary scientific discovery. This project will create a structure-informed framework for discovering how specific parts of proteins influence gene activity. The research will bring together genetic variation, gene expression, protein abundance, protein structure, and experimental perturbation data from large public datasets of cell line models. First, the team will map naturally occurring genetic changes onto three-dimensional protein structures and identify protein regions where changes are consistently linked to changes in gene activity. These regions will then be used to train interpretable Artificial Intelligence models that predict which proteins, and which parts of those proteins, are likely to control specific gene programs. Second, the team will test selected predictions using genome-editing experiments, beginning with the oxidative stress response, a protective system that helps cells respond to damaging conditions. Third, the team will examine how cells control the levels of key proteins that regulate genes, including systems that mark proteins for degradation. By connecting large-scale data analysis with targeted experiments, this work will provide a general strategy for turning complex biological datasets into mechanistic hypotheses that can be tested in the laboratory. The results will advance understanding of gene regulation and provide reusable computational and educational resources for the broader scientific community. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria. NSF Award ID: 2543774 | Program: 01003031DB NSF RESEARCH & RELATED ACTIVIT,01002627DB NSF RESEARCH & RELATED ACTIVIT | Principal Investigator: Elizabeth Brunk | Institution: University of North Carolina at Chapel Hill, CHAPEL HILL, NC | Award Amount: $598,807 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2543774 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2543774.html