Phenomics-enabled prediction of immu no-inflammation drug responses

Immune and inflammatory responses are central to drug efficacy and safety, with dysregulated immune activation underlying adverse events such as cytokine storms, autoimmunity, and chronic inflammation. However, predictive approaches for immune–inf lammation liabilities remain limited, often relying on incomplete omics or cheminformatics datasets. In a project supported by the Sanofi iDEA-Tech award, we combined Sanofi’s therapeutic interest in immune–inflammation with pixlbio’s expertise in highcontent morphological profiling and AI-driven analytics. We developed a domain-specific framework for mechanism-of-action (MoA) and safety prediction of immuno-inflammatory compounds using Phenomics. High-quality, dose-response datasets were generated in multiple immune–inflammation-relevant cell lines, focusing on compounds with well-annotated MoA and safety/toxicity endpoints. Deep learning combined with conformal prediction enabled uncertainty-aware MoA and safety classification, improving predictive accuracy and generalizability. Our fully robotized, iterative workflow demonstrated the feasibility of building high-quality, domain-specific datasets for robust modeling. This collaboration highlights our Phenomics platform as a scalable, AI-driven approach to drug safety and phenotypic screening in immuno-inflammatory drug discovery.