The main interest of the team is the inference and exploitation of molecular networks to extract information from complex omics datasets. We develop methods and apply them in collaboration with biologists and clinicians to the analysis of small, specific tumor cohorts. A strong emphasis is put on the tumor microenvironment and its cellular networks, which we investigate using bulk, single-cell, or spatially resolved data. We are starting to apply these techniques to autoimmune diseases. Network science and machine learning (ML)/AI are the preferred tools for this type of study.
We also develop methods related biomarker discovery. In the field of cancer research, we mainly work on circulating DNA and fragmentomics data using ML/AI. We also employ numerical analysis and Bayesian approaches to model mass spectrometry data and protein turnover in vivo in Alzheimer disease searching for abnormal dynamics.
Axis 1: Molecular network inference (Lead: Ravel & Colinge)
The analysis of large-scale (multi-)omics datasets is often difficult. Contextualizing data using molecular interaction networks facilitates the extraction of reliable information. We have developed a unique framework to infer cellular networks combined with intracellular pathways. Such analyses can be conducted from bulk, single-cell and spatial data alike (Cabello-Aguilar, NAR, 2020; Villemin, NAR, 2023), which we keep extending with AI techniques notably.
We also proposed a new family of statistical models to infer molecular interactions from perturbation data (Borg, Bioinformatics 2023 & bioRxiv 2024). Generalization of this theory is an active topic.
Axis 2: Biomarker discovery through mathematical modeling (Lead: Colinge)
Applying numerical analysis and ML/AI methodology, we are engaged in various biomarker discovery research programs. A first example concerns the search for abnormal protein dynamics in Alzheimer disease (Lehmann, Anal Chem, 2019; Giroux, J Proteome Res, 2024; Lehmann, Bioinformatics, 2024).
Another example is circulating DNA in cancer detection (Tanos, Adv Sci, 2020). Furthermore, we applied notions of spatial statistics and ML to investigate tumor stiffness spatial heterogeneity relation with clinical parameters (Lopez-Crapez, Sci Rep, 2022), or classical ML to characterize the tumoral microenvironment (Benhamouda, Clin Cancer Res, 2022).
Axis 3: Applications of systems biology (Lead: Cornillot & Colinge)
The development and the application of advanced computational methodologies related to functional genomics leads to recurrent collaborations with biologists and clinicians. Typical studies in this category consist of small- to medium-sized tumor cohorts or specific biological models (Alame, Theranostics, 2020 & 2021; Giguelay, Theranostics, 2022; Souche, Endoscopy, 2021; Honda, Theranostics, 2024; Figarol, Nat Comm, 2024; Singh, Nat Microbiol, 2023; Rabia, Front Immunol, 2023). We typically combine data integration with methods of the axes 1 & 2 or existing tools by others.
Located in the Parc Euromédecine, the Institut de Recherche en Cancérologie de Montpellier (U1194) is located on the Val d'Aurelle Campus (ICM, Institut régional du Cancer Montpellier/ Val d'Aurelle).
U1194 Research Centre supported by Inserm, the University of Montpellier and the Institut du Cancer de Montpellier (ICM)
Axis 1: Multi-OMICS integrative projects to characterize resistant or aggressive tumor cellular networks
Owing to previous expertise in data integration over molecular networks, we decided to establish collaboration with pathologists to develop projects aimed at characterizing rare, aggressive tumors for which small numbers of patient samples are available. By means of extensive data integration we try to compensate for the smaller number of replicates available (20 to 30 typically) thus ensuring confidence in the observations. In particular, we are interested in potential targets to kill or resensitize such tumors to chemotherapy. We established a pipeline of bioinformatics procedures and a methodological toolbox to support such projects. This work includes novel methods inferring cell-cell interactions taking place within the tumour microenvironment (TME), for instance in the case where single cell transcriptomes are available.
We also apply our multi-omics framework to colorectal metastases mapping spatial heterogeneity or e.g. to lung adenocarcinoma following tumor development and resistance emergence in a preclinical model model.
Multi-omics projects involve multiple collaborations with IRCM teams (Maraver, Turtoi, Cavaillès, Pèlegrin) as well as national and international collaborators including clinicians.
Axis 2: Modular response analysis (MRA)
While Axis 1 comprises the inference of intra cellular networks by integrating various pathway databases as well as a cell-cell interaction database that we develop internally, it is limited to mapping existing network biology knowledge to fresh data. We hence initiated a dual research axis around perturbations experiments. We develop methods to learn the quantitative coupling, i.e. weighted directional networks, linking a set of molecules of interest that can be transcripts or proteins. The procedure involves performing quantitative measures on the latter molecule activity depending on their nature (expression, activation, etc.) and mathematical modelling. It has the advantage of being generally applicable and to limit the number of required experiments. We implement and develop new methods around MRA (Santra et al., 2018), which we apply to transcriptional networks (collaboration with the Cavaillès Team) and pancreatic cancer kinase receptor signalling and dynamics upon antibody therapy (collaboration with the Pèlegrin Team).
Axis 3: Computational proteomics and mass spectrometry
The team has a strong expertise in many topics of computational proteomics (identification, quantitation, PTMs, AP-MS data filtering and modelling) and continues collaboration with several teams locally and internationally. Our main project at the time is the assembly of computational and mathematical models for a ground-breaking proteomics technology (proteome wide SILK, pwSILK for short) that allows determining protein turnover in human in vivo and proteome wide (collaboration with the Lehmann and Hirtz Teams and at IRMB, Montpellier). We established automated and fully parallelized pipelines for processing pwSILK complex data, which include a novel differential equation-based model coupled with statistical procedures. We showed that our new mathematical model is able to extract degradation rates from data generated by us and others, following different protocols, reproducibly and consistently. Recently, we could determine the degradation rates of ~200 proteins in ventricular CSF in vivo, an unprecedented result in the field since such data were available for only 6 proteins so far. More projects are coming combining multiple biofluids with the ambition to explore a new class of biomarkers related to protein clearance abnormalities or biological barriers defects. We also have plans to apply pwSILK to cancer metabolism.
Institut de Recherche en Cancérologie de