Kernel-Based Microfluidic Constriction Assay for Tumor Sample Identification | JoVE Visualize

Area of Science:

Biophysics
Cell Biology
Medical Diagnostics

Background:

Distinguishing malignant from nonmalignant breast cells is crucial for effective cancer diagnosis and treatment.
Current methods for analyzing cell biomechanics can be time-consuming and require large sample volumes.
Understanding cell deformability differences is key to identifying cancerous cells.

Purpose of the Study:

To develop and validate a high-throughput microfluidic device for distinguishing human breast cancer cell lines and primary tumor cells from normal cells.
To investigate the utility of cell velocity increments through multiconstriction channels as a marker for malignancy.
To apply kernel learning methods for analyzing high-dimensional biomechanical data.

Main Methods:

Utilized a high-throughput multiconstriction microfluidic channel device to measure cell velocity increments.
Employed machine learning algorithms including Ridge, nonnegative garrote on kernel machine (NGK), and Lasso for data analysis.
Analyzed high-dimensional variables such as cell size, velocity, and velocity increments.
Applied 10-fold cross-validation in kernel learning methods to assess group differences.

Main Results:

Successfully distinguished human breast cancer cell lines (MDA-MB-231, HCC-1806, MCF-7) from immortalized breast cells (MCF-10A) with 81-85% confidence at 50-70 cells/min.
Differentiated primary breast cancer cells from patient-matched adjacent normal cells with 70.07%-75.96% accuracy using the NGK method.
Observed differences in velocity increments through microfluidic channels correlate with cell deformability, distinguishing malignant from nonmalignant cells.

Conclusions:

The multiconstriction microfluidic device combined with kernel learning offers a high-throughput method for analyzing cell biomechanics.
This approach can effectively perturb and analyze cells from small primary tumor biopsy samples.
The developed biomechanical velocity signatures show promise as a novel marker for malignancy identification and risk assessment.