Machine Learning in Life Sciences Laboratory

We develop and apply machine learning and deep learning models, together with explainable AI techniques, to interpret tumor biology from bulk and single-cell transcriptomics. Our focus is triple-negative breast cancer and other hard-to-treat subtypes, where interpretable models help identify therapeutic targets, characterize resistance mechanisms, and guide translational decisions.

We develop computational workflows for drug repurposing based on docking, pharmacogenomics, transcriptomic signatures, and structure-based target prioritization. Within the RePo-SUDOE transnational network we operate three public platforms: RePo-EDU for educational resources and training materials, RePo-VRE as a virtual research environment for drug repurposing, and RePo-BOT as an automated assistant for project queries. In PRECISION we apply graph learning to prioritize novel targets and repositioning opportunities in oncology.

We build ML pipelines to analyze 16S and shotgun microbiome data in inflammatory bowel disease, type 1 and type 2 diabetes, preterm birth, and vaginal microbiota. In parallel, we study antimicrobial resistance in nosocomial pathogens, with ongoing work on Acinetobacter and Klebsiella pneumoniae genomics, surveillance, and clinical outcome prediction.

We combine genomics, transcriptomics, proteomics, and single-cell data to study molecular interactions across biological layers. We also integrate medical imaging and protein structural information into the same modeling pipelines, improving diagnosis and prognostic modeling in complex diseases.