OET Cell Signature: Cells Discrimination and Drug Response Evaluation with Opto-Electronic Tweezers and Machine Learning Algorithms

Cell responses to varying electric fields can reveal insights on cell biology, with important implications in basic and pharmaceutical research. In this work, Opto-Electronic Tweezers (OET) are used for inducing label-free cell translations at multiple frequencies and, in combination with time-lapse measurements and machine learning, to derive characterizations of single cells as well as of cell populations. A customized polymethymetacrilate (PMMA) chip with ITO substrates and an a-Si layer was designed for OET-based manipulation of cells and integrated with an inverted microscope. We obtained OET cell signatures as kinematic and dynamic spectra depicting the translational responses induced via the OET at increasing frequencies. Starting from the OET signatures, we used machine-learning algorithms to find the best set of descriptors, which enabled automatic discrimination of cell types and treatment conditions. After calibration with polystyrene beads, experiments were performed on three biological case studies, involving 1) the discrimination of cell types among U937 human leukaemia cells, PC-3 human prostate cancer cells, and HaCaT human immortalized keratinocytes; the evaluation of the effects of the chemotherapeutic agent etoposide 2) on U937 cells at different concentrations; and 3) on PC3 at different exposure times. The accuracy values of 99.05% (2.13%), 96.11% (5.41%) and 98.89% (2.48%), obtained for the three experimental scenarios, respectively, show that multiple levels of dielectric information can be extracted from the OET cell signatures and combined together to improve cells discrimination and drug response evaluation.