Jiahua Rao

We present RiboFlow, a flow-matching based model that co-designs RNA structures and sequences with high binding affinity to target molecules.

We present H3-DDG, a model that predicts binding free energy changes by modeling higher-order many-body interactions, achieving state-of-the-art performance in protein mutational effect prediction.

We present GLARE, a reinforced active learning framework that improves virtual screening efficiency by dynamically optimizing molecular selection while balancing chemical diversity, biological relevance, and computational constraints.