Computational biology & bioinformatics

The group develops computational and statistical methods for extracting biological meaning from large, noisy, high-dimensional measurements — gene-expression profiles, mutation data, and other genomic and proteomic readouts. Drawing on ideas from statistical physics, the work treats clustering, dimensionality reduction, and classification as problems of finding robust structure in data where the number of measured features vastly exceeds the number of samples.

A central thread is the analysis of cancer at the level of biological pathways rather than individual genes. By aggregating molecular measurements onto known pathways and protein interaction networks, the group aims to obtain stable, interpretable signatures that distinguish disease subtypes and inform personalized characterization of tumors. The motivation is practical as well as conceptual: pathway-level representations tend to be more reproducible across cohorts than single-gene markers, and they connect statistical patterns to mechanism.