Diffusion model predicts the geometry of actin cytoskeleton from cell morphology

Area of Science:

• Cell Biology
• Biophysics
• Computational Biology

Background:

• Cell morphology is linked to actin cytoskeleton dynamics, specifically stress fibers.
• Stress fibers are crucial for cell shape, mechanotransduction, and physiological processes.
• Geometric principles connecting cell shape and stress fiber organization remain largely unexplored.

Purpose of the Study:

• To develop a machine learning system to predict stress fiber distribution from cell morphology.
• To investigate potential geometric principles governing the relationship between cell shape and actin cytoskeleton organization.
• To enable virtual experiments for exploring subcellular structure configurations.

Main Methods:

• Utilized a diffusion model-based machine learning system.
• Trained the system on corresponding cell shape and stress fiber datasets.
• Generated stress fiber images from cell shape inputs.

Main Results:

• The machine learning system accurately predicts stress fiber distribution and alignment.
• Predicted distributions align well with experimental data.
• Identified a correlation between cell aspect ratio and stress fiber localization (e.g., edge localization in narrow cells, homogeneous distribution in wide cells).

Conclusions:

• The developed system effectively converts cell shape to stress fiber patterns.
• This approach facilitates the discovery of hidden geometric principles in subcellular organization.
• Cell shape geometric constraints significantly influence actin stress fiber arrangement.