Xiaoning Qi

The paper introduces CEE, a framework to assess genetic perturbation target identification by eliminating evaluation artefacts, and finds simple methods outperform complex models.

This paper develops a language-model agent for hit-to-lead optimization that outperforms existing methods on four kinase targets in a computational benchmark.

CellOS is a multi-view foundation model trained on single-cell transcriptomes that outperforms existing models in cell-state annotation, batch integration, and perturbation-response prediction.