Computational cancer biology

Tumors form highly complex structures comprising many different cell types like cancer and immune cells. In our research, we develop novel computational biology and immuno-informatics tools to better understand the interactions between immune and cancer cells.

Computational analysis of HLA peptidomics data and predictions of T cell neo-epitopes

In-depth and unbiased identification of peptides presented on HLA molecules with Mass Spectrometry provides a unique opportunity to collect very large HLA ligand datasets that can inform us about the binding properties of HLA molecules and can be used for training HLA ligand predictors. Through the development of novel machine learning algorithms, our lab has been the first to show how to deconvolute pooled HLA peptidomics data (i.e., coming from samples expressing up to 6 HLA-I alleles) and how to use these data to improve predictions of neo-antigens [Bassani-Sternberg and Gfeller 2016, Bassani-Sternberg et al. 2017]. Currently we are expanding these analyses to different types of HLA-I ligands and to HLA-II molecules [Racle et al. 2019].

Computational analysis of immune infiltrations in tumors

Tumors are composed of various cell types. Unfortunately cancer genomics studies are often restricted to bulk tumors. Our lab has recently developed a novel computational approach to simultaneously Estimate the Proportion of Immune and Cancer cells (EPIC) from bulk tumor gene expression data that can quantitatively predict the fraction of all major immune cell types, as well as cancer cells [Racle et al. 2017]. This will expand the scope of prospective analyses of immune infiltrations using gene expression profiling and enable retrospective analyses of thousands of cancer genomics datasets from human patients. In parallel, we are actively working on single-cell RNA-Seq data analysis for cancer and immunology, to explore cell type heterogeneity in a fully unbiased and marker-free approach. Among else, our work has led to the first molecular evidence and characterization of NK-like cells in non-mammalian species [Carmona et al. 2017].

Our computational tools are made available on the github of the lab : https://github.com/GfellerLab

• Julien Schmidt, Angela R. Smith, Morgane Magnin, Julien Racle, Jason R. Devlin, Sara Bobisse, Julien Cesbron, Victor Bonnet, Santiago J. Carmona, Florian Huber, Giovanni Ciriello, Daniel E. Speiser, Michal Bassani-Sternberg, George Coukos, Brian M. Baker, Alexandre Harari, David Gfeller, Prediction of neo-epitope immunogenicity reveals TCR recognition determinants and provides insight into immunoediting, Cell Reports Medicine, (2021).

• Racle J, Michaux J, Rockinger GA, Arnaud M, Bobisse S, Chong C, Guillaume P, Coukos G, Harari A, Jandus C, Bassani-Sternberg M, Gfeller D, Robust prediction of HLA class II epitopes by deep motif deconvolution of immunopeptidomes, Nat Biotech (2019).

• Guillaume P, Picaud S, Baumgaertner P, Montandon N, Schmidt J, Speiser DE, Coukos G, Bassani-Sternberg M, Fillipakopoulos P, Gfeller D, The C-terminal extension landscape of naturally presented HLA-I ligands, PNAS, 115, 5083 (2018).
• Racle J, De Jonge K, Baumgartner P, Speiser D, Gfeller D. Simultaneous enumeration of cancer and immune cell types from bulk tumor gene expression data, eLife, 6:e26476 (2017).
• Bassani-Sternberg M, Chong C, Guillaume P, Solleder M, Pak HS, Gannon PO, Kandalft LE, Coukos G, Gfeller D. Deciphering HLA-I motifs across HLA peptidomes improves neo-antigen predictions and identifies allostery regulating HLA specificity, PLoS Comp Bio, 13:e1005725 (2017).

• Sinergia
• Swiss National Science Foundation
• Swiss Cancer League
• Systems X