In-flow Motion Dynamics for Mechanical-Based Clustering of Cells Under Different Compression Conditions

Cell mechanical properties are powerful biomarkers for label-free phenotyping. However, especially in cancer, a large heterogeneity of deformability properties is present and measuring such wide variety is still challenging. Thus, a simple, versatile, and cost-effective method for detection of variable rheological/mechanical cell properties is needed. Here, we present a reduced set of motion parameters as the local cell velocity, the lateral equilibrium position and the orientation angle capable to recognize a wide range of cell mechanical properties, in microfluidics. By ranging the applied in-flow viscoelastic compression forces, we deform cells inducing different deformation-dependent dynamics (rolling, tumbling, swinging, tank-treading). From such motion characteristics, we identify cell clusters with unsupervised machine learning approaches, discerning different cell deformability levels after multiple levels of applied forces. Thus, our approach represents as a label-free opportunity to characterize and classify different cell classes and subclasses under different mechanical conditions.