Seminar| Institute of Mathematical Sciences
Time: Friday, April 24th, 2026,13:30-14:30
Location: IMS, RS408
Speaker: Shuhui Wang, Université Paris Dauphine-PSL
Abstract: My work aims to significantly improve cancer drug discovery and treatment within the context of precision medicine by addressing the limitations of traditional chemotherapy, cancer's genetic complexity, and the high cost of comprehensive drug screening.
I tried to tackle the drug discovery challenge by decomposing it into three fundamental, interconnected subproblems, each addressed by a dedicated Deep Learning solution:
1. Gene–Gene Interactions – understanding how combinations of mutations shape cellular phenotypes. We proposed D-LIM, an interpretable neural network to model the genotype–phenotype–fitness landscape to gain mechanistic insights.
2. Gene–Drug Interactions – mapping how drugs perturb gene expression and alter cell states. We developed DORA (DOse Response Autoencoder) to predict transcriptomic changes and linking them to cancer cell viability to identify actionable biomarkers.
3. Drug–Drug Interactions – identifying synergistic effects among drug combinations. We prioritized synergistic drug combinations by proposing an active learning workflow to bridge computational prediction with targeted experimental validation.
By connecting genotype, phenotype, and drug response, my work contributes to the development of interpretable AI frameworks for precision oncology and more efficient cancer drug design.