Multi-scale graph neural networks for modeling and designing RNA sequence-structure-function

Faculty Mentors: Julius Lucks (Northwestern University) & Risi Kondor (University of Chicago)

This project seeks to revolutionize our understanding of RNA by modeling the complex relationships between its sequence, structure, and function. By leveraging the mathematical framework of graph neural networks (GNNs), the researchers represent RNA molecules as graphs where nucleotides act as nodes and structural features—such as base-pairing and motifs—act as edges. Building on successful validations using riboswitch dynamics, the team is extending this approach to create multi-scale generative models. This innovation will move the field beyond simple functional prediction, enabling the de novo design of novel RNA sequences with precise, tailored functional properties.