Recent advances in single-cell and spatial omics technologies have opened a unique opportunity to explore biological systems at an unprecedented resolution and scale. Our lab develops novel computational and statistical methods to translate the complex data generated by these technologies into biological understanding. We have created several widely-used computational methods tackling multiple challenges in single-cell data science including integration, celltype classification, reference atlas mapping, gene signature scoring, and T cell clonal analyses.

The tumor microenvironment (TME) is the complex niche surrounding tumors, composed of various cell types and molecules that influence tumor growth and treatment responses. Beyond cancer cells, it contains a striking diversity of immune cells, fibroblasts, endothelial cells, and other tissue-specific cells, all of which evolve with cancer progression. The components of the TME and their complex interactions determine to a large extent disease progression and response to therapies. Deep characterization of the TME holds the potential to reveal new therapeutic targets and mechanisms of therapy resistance. Single-cell and spatial transcriptomics technologies provide the most powerful tools to resolve cell heterogeneity in complex tissues. Thanks to our cutting-edge computational methods, we can analyze single-cell transcriptomics data at scale, across thousands of patients and millions of cells, to dissect the TME composition at high resolution and characterize patient-to-patient variation within and across cancer types. We aim to identify recurrent TME patterns across patients that are associated with clinical outcome to infer resistance mechanisms and predict new therapeutic targets.

Therapies that modulate the immune system, such as immune checkpoint blockade, are revolutionizing the landscape of cancer treatment. Yet, only a fraction of patients successfully respond to these therapies across cancer types. Profiling the immune landscape presents a great potential for the discovery of biomarkers of therapy response. In particular, blood samples emerge as accessible, minimally invasive, and inexpensive windows on the immune status of an individual that could be routinely measured in the clinic. We are developing statistical and machine learning models to predict disease state and response to therapy in cancer patients based on single-cell transcriptomics data from tumor biopsies and blood samples. 

T cells are crucial players in the adaptive immune response with the capacity to recognize and eliminate infected and malignant cells and to mediate autoimmune disease. Depending on the immunological context, T cells acquire distinct differentiation states with different functions – e.g. Th1 vs Tfh CD4+ T cell states. A T cell state that might be beneficial in one immunological context (e.g. viral infection) might be less beneficial or detrimental in another one (e.g. autoimmune disease). Thus, it is critical to understand what is the full spectrum of T cell differentiation states, their functions and role in disease outcome. Moreover, in vitro T cell manipulation – e.g. by metabolic reprogramming or genetic engineering – can expand the landscape of T cell states and functionalities observed in vivo. These are key for developing successful T cell-based therapies, such as CARs, TCR-engineered T cells and TIL therapies. In collaboration with our experimental partners, we are conducting meta-analyses of single-cell omics data across individuals, tissues, diseasesand in vitro conditions in mouse and human to characterize the landscape of T cell states, their potential functions, their couplings with T-cell receptor sequence and antigen specificity, and their associations with disease outcome.