Limeng Pu

Seq2Karyotype (S2K) enables in-silico karyotyping from whole-genome sequencing data, revealing intratumoral heterogeneity and evolutionary trajectories in cancer samples.

GraphSite classifies ligand-binding sites using a deep graph neural network, achieving 81.28% test accuracy on 14 classes.

GraphGR predicts the effect of kinase inhibitors on cancer cell growth using a graph neural network that integrates heterogeneous data, achieving high accuracy on unseen data.