Curvature-Guided Geometric Representation for Protein-Ligand Binding Affinity Prediction

Researchers have proposed two new methods, RicciBind and CPES, for predicting protein-ligand binding affinity, a critical task in drug discovery.

RicciBind, described in a paper on arXiv^[1], integrates curvature-guided hierarchical structure learning with optimal transport-based cross-domain alignment to model molecular interactions. It leverages Ricci curvature to capture local interaction tightness within molecular structures and aligns protein and ligand clusters across heterogeneous domains under geometric constraints. This approach substantially improves both the accuracy and interpretability of binding affinity prediction, according to the researchers. Extensive experiments demonstrate that RicciBind achieved superior predictive performance and generalization across PLA benchmarks and virtual screening tasks^[1]. Another method, CPES, proposed in a separate paper on arXiv^[2], uses curvature representations to model conformational flexibility and spectral cross-attention to compare the unbound ligand and protein with the bound complex. CPES fuses curvature-informed dynamic representations with static interaction representations for affinity regression, achieving improved predictive performance and offering physical interpretability^[2].