My research focuses on statistical learning and artificial intelligence methods for the analysis of single-cell and spatial omics data. I am particularly interested in three central challenges: separating biological from technical variability, scaling inference to increasingly high-dimensional and heterogeneous data, and maintaining rigorous statistical risk control as resolution increases.
Statistical inference for high-dimensional biological data
Statistical hypothesis testing remains central to reproducible biological discovery. I develop kernel-based and non-parametric testing methods for complex high-dimensional data, with a particular interest in RKHS methods, general experimental designs, post-clustering inference and scalable procedures with controlled statistical risk.
Spatial single-cell data
Spatial transcriptomics technologies such as Visium and Xenium add tissue organization to molecular measurements. I develop methods based on spatial statistics, point processes and kernel testing to characterize spatial variability, compare molecular distributions and study interactions between cell populations and tissue structures.
Explainable AI and sensitivity analysis
As models become more flexible, understanding why they detect a biological signal becomes essential. I develop sensitivity analysis and explainable AI approaches to identify the cells, genes and molecular features that drive statistical or predictive differences.
Representation and integration
Single-cell experiments increasingly combine several molecular modalities, individuals and biological scales. I develop statistical learning methods for dimension reduction, representation and integration of heterogeneous omics data, including approaches based on probabilistic modeling and optimal transport.
Biological applications
My methodological work is developed in close collaboration with experimental and computational biologists. Applications are primarily in developmental biology and cancer biology, where single-cell and spatial measurements provide new ways to study cell states, tissue organization and disease heterogeneity.
Previous research
My earlier work includes functional data analysis, point-process modeling for genomic data, population genomics, statistical genomics and the spatial program of DNA replication. These projects continue to inform my current methodological work, particularly in functional and spatial statistics.