A Machine Learning Approach to Predict Cellular Mechanical Stresses in Response to Chemical Perturbation

Mechanical stresses generated at the cell-cell level and cell-substrate level have been suggested to be important in a host of physiological and pathological processes. However, the influence various chemical compounds have on the mechanical stresses mentioned above is poorly understood, hindering the discovery of novel therapeutics and representing a barrier in the field. To overcome this barrier, we implemented two machine learning (ML) Models: Stepwise Linear Regression (SLR) and Quadratic Support Vector Machine (QSVM) to predict the dose-dependent response of tractions and intercellular stresses to chemical perturbation. We used traction and intercellular stress experimental data gathered from 0.2 μg/mL and 2 μg/mL drug concentrations along with cell morphological properties to train our model. To demonstrate the predictive capability of our ML model we predicted tractions and intercellular stresses in response to 0 μg/ml & 1 μg/ml drug concentrations. Results revealed the QSVM model to best predict intercellular stresses, while SLR best predicted tractions. Author Summary The ML framework we present here can be used to predict the mechanical response of any anchorage-dependent cell type to any chemical perturbation. The proposed ML can directly predict the intercellular stresses or tractions as a function of drug dosage and/or monolayer/cell coverage area which could potentially reduce the experimental time on studying the mechanics of cells to external chemicals or mechanical constraints. We believe our findings could be helpful in accelerating drug discovery and increase our understanding in the role of cellular stresses during disease progression.