Integrative genomics

We develop tools to analyze and combine newly generated and existing open-source data. Genome-wide transcriptional profiling, using spatial or single-cell transcriptomics, can help to identify candidate regulators and drivers of the disease under investigation. Our methodological research emphasizes on gene regulatory networks and network analysis to prioritize key regulators that govern developmental and pathological processes. In order to study mood disorders, we use integration of multi-omics data (mRNA-seq, miRNA-seq, methylomes, and GWAS) with phenotype data from patients to identify candidate biomarkers and to help to stratify patient groups. Beyond a better understanding of molecular drivers of neurodevelopment and neuropsychiatric disorders, we aim at providing new candidate targets for drug discovery and repurposing. To this purpose, we currently study the use of machine learning methods to repurpose drugs based on their transcriptional impact.

VS360⁰: Integrated AI-assisted virtual screening pipelines and database mining to assist early stages drug discovery and foster actionable knowledge (Bruno Villoutreix)

 The aim of virtual screening (VS) is to identify
bioactive compounds (small molecules, short peptides…) through computational means by using as input the 3D structure of a putative protein target (structure-based VS) and/or known
bioactive molecules against this selected target (ligand-based VS). In VS, a large number of
molecules (approved drugs, already synthesized compounds, virtual compounds) are ranked according to their likelihood of being bioactive using different metrics. The top fraction of the resulting predicted list can then be tested experimentally. Modified VS algorithms can also be used to predict off-targets and anti-targets, drug combination, multi-target drugs… While such computations assist decision making, in real life scenario, many other information are needed so as to get a panoramic view about the investigated drug-target-disease axis (e.g., projection of the compounds onto the human interactome, onto disease pathways, consideration of ADMET properties…). We develop VS360^O, a tool that combines LBVS and SBVS and various data mining approaches so as to generate a comprehensive profile around the investigated system (i.e., a 360^o view with integrated experimental and predicted data). VS360^O can be used for drug repurposing, to discover novel high quality compounds and novel targets/pathways/mechanisms. The tool can be easily combined with experimental methods and should definitively contribute to de-risking therapeutic development in an academic setting.