ID-4-IN-6094 A High Content Screening Platform for Phenotypic Drug Discovery: Automated Image Analysis of Complex Cell-Based Model Systems for Oncology and Toxicology

Complex diseases such as human cancers require complex, physiologically relevant model systems that allow truly informative testing of pharmacological compounds and biosimilars in vitro. Traditionally, most model systems routinely used in drug discovery represent highly reductionist approaches, and fail to address many key aspects of genuine tissue architecture (histology). However, organotypic cell- and tissue culture models that more faithfully mimic the tumor microenvironment (TME), grapidly gain acceptance. These systems aim to recapitulate the heterotypic interaction of multiple cell types (e.g. tumor and endothelial or stroma cells) and physiological extracellular matrices (ECM, basement membranes). The main challenges are to develop a balanced approach between sample throughput and biological relevance, to provide better but accessible and affordable in vitro tools to replace animal testing and predict human risk assessment. Although such assays are increasingly being utilized for prediction of drug efficacy versus organ-specific toxicity, they often remain poorly characterized. Additional features that require more attention are the dynamics of stromal and tumor cells (tissue homeostasis versus invasion and metastasis), tumor cell plasticity (the capacity to change cellular phenotypes), and tumor heterogeneity (the complex and shifting clonal composition of tumor tissues). Very few screening platforms for drug discovery systematically address these key elements of cancer biology. In this lecture, we will introduce our approach that allows real-time, live cell monitoring of cellular dynamics using diverse microscopic techniques, automated image analyses, combined with machine learning. The major focus will be on machine vision solutions, developed specifically for tracking of dynamic stromal and tumor cell motility. This is combined with unsupervised, statistical methods to capture the intrinsic heterogeneity of tumor tissues. The final goal is to provide a high content screening platform for phenotypic drug discovery with a significantly improved experimental throughput - without sacrificing biological relevance.