Our group develops statistical methods and tools- particularly probabilistic and Bayesian approaches- for novel single and spatial technologies.

One of our major accomplishments in this field is our contribution to MOSAIC (Multi-Omics Spatial Atlas in Cancer) - a multi‑center clinical project aimed at comprehensively characterizing thousands of tumor samples with spatial and single‑cell technologies across diverse cancer types. Whitin this international initiative, we contributed our expertise in advanced data science to help build the world’s largest spatially-resolved omics reference database in oncology (MOSAIC: Intra-tumoral heterogeneity characterization through large-scale spatial and cell-resolved multi-omics profiling | bioRxiv; Dong et al, 2025).

In recent years, one of our most significant achievements has been the development of the first statistical model to account for the bimodal gene expression patterns commonly observed in single‑cell data. This model is implemented in the MAST package, which has since become one of the most widely cited methods for differential gene expression analysis in scRNA‑seq studies.

We also introduced one of the first models for the analysis of spatial transcriptomics that enables near single-cell resolution analysis of 10X Visium data. This method is implemented in the open-source BayesSpace package. Both MAST and BayesSpace are available from the Bioconductor project. More recently, we have developed additional methodologies for spatial transcriptomics, including tools for preprocessing, segmentation, machine learning, and AI-based analysis. We have also made key contributions to the Bioconductor ecosystem through the OSTA book (https://bioconductor.org/books/release/OSTA/).

Using single‑cell and spatial computational analyses, our group contributes to numerous studies aimed at characterizing the tumor microenvironment (TME) —including immune cells—to better understand mechanisms of response and relapse to immunotherapy.

One example is our participation in IMMUcan (Integrated iMMUnoprofiling of large adaptive CANcer patient cohorts), a European consortium focused on leveraging multi‑omics data to uncover how the TME drives cancer progression. To support this effort, we provide our expertise in data integration and pan‑cancer analyses to identify shared spatial patterns of the TME across cancer types. We also apply emerging spatial omics methods to the consortium’s spatial proteomics data and collaborate closely with other groups to integrate multi‑omics data across different cancers.

In previous years, we have developed and applied computational tools to identify novel biomarkers of responses to anti-PD1 therapy (Greene et al 2021) and escape mechanisms to adoptive T-cell therapies (Paulson et al 2018, Chapuis et al 2019, Lahman et al. 2022).