Systematic AI-assisted screening of the cadhesome to map epithelial monolayer mechanics.

Cadherin-mediated adhesions serve as key mechanical and signaling hubs in epithelial tissues, linking the actin cytoskeleton of adjacent cells. Their disruption is a hallmark of cancer progression. The "cadhesome" network comprises over 170 proteins involved in cadherin-mediated adhesion and force transmission, yet its complexity hampers functional understanding. We developed a high-throughput platform combining gene silencing, imaging, and AI-based analysis to profile the role of each cadhesome component in monolayer formation and mechanical integrity. Using EpH4 epithelial cells, we analyzed phenotypes under vehicle and nocodazole-challenge conditions. Machine learning enabled classification of monolayer disruption, junctional organization, and contractile state. Beyond confirming known mechanotransduction hubs centered on E-cadherin, EGFR, and RAC1, our approach systematically uncovered candidate regulators of monolayer contractile state and stress adaptation, identified condition-specific roles of poorly characterized proteins, and organized them into annotated mechanobiological subnetworks that serve as a basis for hypothesis generation. Presented as a prioritized discovery resource, this work establishes a scalable strategy to decode mechano-molecular networks and provides a blueprint for hypothesis-driven investigation of epithelial mechanics with potential translational relevance.