Machine Learning of Cell Mechanobiology Experimental Results to Determine Metastatic Risk

The main risk for cancer patients is metastasis of invasive tumors to distant sites. Metastases form when cells are shed from the primary tumor, then forcefully migrate to new sites, and create new cancerous tumors. The invasive capacity of cells is related to and determined by their mechanobiology, specifically the interaction of the cancer cells with their environment. The Weihs lab have shown that a subset of invasive cancer cells will indent a synthetic, impenetrable, physiological-stiffness polyacrylamide gel, while benign cells do not indent. The novel in vitro assay determines the mechanical invasiveness of cells, which, as the Weihs lab have recently shown, directly agrees with the metastatic risk of cancer cells in vivo. The mechanical invasiveness, or indentation capacity of the cell sample is defined through the percent of indenting cells (the invasive subset) and their attained indentation depths. We have utilized these experimental results to develop supervised machine learning modules to identify and predict the metastatic potential of a sample. Concurrently, it is important to evaluate the mechanical forces applied by invasive cells as well as their mechanisms, as those may provide targets for interventions. Hence, we have developed finite element simulations of the indentations to estimate the forces applied by group of cells in varying experimental configurations.