Prediction of Piconewton Receptor Tension Images using Deep Learning.

Area of Science:

• Cellular biology
• Biophysics
• Bioengineering

Background:

• Piconewton (pN) receptor forces are crucial for cellular functions but difficult to measure.
• Molecular tension probes (MTPs) offer a way to measure these forces but have limitations in complex cellular environments.

Purpose of the Study:

• To develop an AI-driven method, Tension Deep Learning (TensionDL), for predicting pN receptor tension maps.
• To overcome the limitations of MTPs in measuring cellular forces, especially in heterogeneous environments.

Main Methods:

• Utilized convolutional neural networks and image-to-image translation.
• Trained the model using cell morphology and vinculin images.
• Validated TensionDL across different cell types and substrate stiffnesses.

Main Results:

• TensionDL accurately predicts pN receptor tension maps at subcellular and cellular levels.
• The model demonstrates robustness across various experimental conditions.
• Enabled semi-quantitative predictions of cell mechanical output and long-term tension mapping.

Conclusions:

• TensionDL provides a powerful, non-invasive tool for mapping pN receptor forces in cells.
• The AI approach enhances the study of cell mechanics in complex and heterogeneous biological systems.
• Facilitates a deeper understanding of how cellular forces influence biological processes.