Rahmatullah Roche

ProRNA3D-single predicts protein-RNA complex structures from single-sequence inputs, outperforming existing methods with limited evolutionary data.

EquiPNAS improves protein-nucleic acid binding site prediction using a pLM-informed equivariant graph neural network, outperforming existing methods on multiple datasets.

EquiPPIS predicts protein-protein interaction sites using E(3) equivariant graph neural networks, outperforming existing methods with predicted AlphaFold2 structures.

PIQLE estimates protein-protein interface quality using deep graph learning, outperforming existing methods by incorporating multimeric geometries and evolutionary information.

rrQNet estimates contact map quality using deep learning by reconciling evolutionary context with predicted residue contacts, showing accurate quality assessment across contact predictors.

DisCovER improves protein threading by integrating inter-residue distance, orientation, and neighborhood effects in template alignment.

This paper introduces DConStruct, a new hierarchical structure modeling method for ab initio protein folding that outperforms CNS in accuracy using distance and contact maps for soluble and membrane proteins.