Beyond the Pattern: Bridging Multimodal Data and Mech anistic AI for Clinical Translation

Abstract

Despite continuous advances in deep learning, most AI systems remain limited to pattern recognition, with little ability to capture underlying mechanisms or provide interpretable reasoning, creating major challenges in biomedical applications involving complex, structured, and heterogeneous data. In this seminar, Professor Occhipinti will show how integrating multimodal data (including spatial omics and imaging datasets) with mechanistically informed AI models can reveal novel functional signatures. Using breast cancer as a paradigm, she will demonstrate how these integrative AI strategies can infer latent system states and map the regulatory principles governing complex biological systems, highlighting their potential for clinical translation.