Deep Learning-Based Prediction of Individual Cell α-Dispersion Capacitance from Morphological Features.

Area of Science:

• Biophysics
• Cell Biology
• Computational Biology

Background:

• Cellular membrane electrical properties offer insights into cell states and cancer diagnostics via epidermal growth factor receptor (EGFR) expression.
• Morphological changes during observation confound electrical measurements, masking true biological responses to epidermal growth factor (EGF).

Purpose of the Study:

• To develop a deep learning method to computationally link cellular morphology and electrical properties.
• To correct for morphology-induced measurement errors in electrical analysis of cells.

Main Methods:

• Combined optical trapping and capacitance measurements on HeLa cells under DPBS and EGF stimulation.
• Developed a convolutional neural network (CNN) to predict capacitance spectra from morphological images.

Main Results:

• CNN accurately predicted capacitance spectra (0.1-2 kHz) from morphological images (<10% error at 0.1-0.8 kHz).
• Method effectively isolated true biological responses by subtracting morphology-dependent capacitance components.
• Demonstrated robust prediction across diverse cell morphologies and experimental conditions.

Conclusions:

• The developed deep learning approach provides a computational framework for correcting morphology-induced errors in electrical measurements.
• This significantly enhances the precision and reliability of EGFR-based cancer diagnostics.