Automated\nBio-AFM Generation of Large Mechanome Data\nSet and Their Analysis by Machine Learning to Classify Cancerous Cell\nLines

Mechanobiological\nmeasurements have the potential to discriminate\nhealthy cells from pathological cells. However, a technology frequently\nused to measure these properties, i.e., atomic force microscopy (AFM),\nsuffers from its low output and lack of standardization. In this work,\nwe have optimized AFM mechanical measurement on cell populations and\ndeveloped a technology combining cell patterning and AFM automation\nthat has the potential to record data on hundreds of cells (956 cells\nmeasured for publication). On each cell, 16 force curves (FCs) and\nseven features/FC, constituting the mechanome, were calculated. All\nof the FCs were then classified using machine learning tools with\na statistical approach based on a fuzzy logic algorithm, trained to\ndiscriminate between nonmalignant and cancerous cells (training base,\nup to 120 cells/cell line). The proof of concept was first made on\nprostate nonmalignant (RWPE-1) and cancerous cell lines (PC3-GFP),\nthen on nonmalignant (Hs 895.Sk) and cancerous (Hs 895.T) skin fibroblast\ncell lines, and demonstrated the ability of our method to classify\ncorrectly 73% of the cells (194 cells in the database/cell line) despite\nthe very high degree of similarity of the whole set of measurements\n(79–100% similarity).