Multiparameter mechanical and morphometric screening of cells.

Area of Science:

• Biophysics
• Cell Biology
• Machine Learning

Background:

• Traditional cell phenotyping often relies on labeling, which can alter cell properties.
• There is a need for rapid, label-free methods to analyze cell physical characteristics for accurate classification.

Purpose of the Study:

• To develop and validate a label-free method for rapid cell phenotyping and classification based on intrinsic biophysical properties.
• To demonstrate the utility of this method for discriminating cell types, states, and viability.

Main Methods:

• Extracting 15 biophysical parameters from cells deforming in a microfluidic stretching flow field using high-speed microscopy.
• Applying machine learning algorithms to classify cells based on extracted rheological and morphological properties.

Main Results:

• Achieved over 95% accuracy in classifying cell pluripotency using the full 15-dimensional dataset.
• Demonstrated successful classification of cell viability, drug screening responses, and detection of malignant cells in mixed samples.
• Identified that a subset of parameters is sufficient for maximum classification accuracy, though informative subsets may vary by cell type.

Conclusions:

• This label-free assay leverages intrinsic cell biophysics for robust cell state identification.
• The method offers a label-free alternative to flow cytometry and provides novel intracellular metrics not feasible with labeled approaches.