Rapid prediction of lab-grown tissue properties using deep learning.

Area of Science:

• Biophysics
• Cell Biology
• Computational Biology

Background:

• Cell-extracellular matrix interactions are crucial for tissue self-organization.
• Mechanobiology plays a key role in the behavior of cell-laden hydrogels.

Purpose of the Study:

• To demonstrate the use of machine learning (ML) for predicting mechanobiology in cell-laden hydrogel self-organization.
• To develop an automated method for generating diverse mould designs for hydrogel growth.

Main Methods:

• Automated generation of tethered mould designs with varying symmetries.
• Creation of a large dataset (N=6400) using biophysical simulations (contractile network dipole orientation model).
• Training a pix2pix deep learning model with simulation data and testing on unseen cases.

Main Results:

• The ML model achieved excellent predictive accuracy compared to biophysical simulations.
• The ML approach demonstrated significantly faster computation times than traditional biophysical methods.
• The study validates the potential of ML for high-throughput design in tissue engineering.

Conclusions:

• Machine learning offers a powerful and efficient tool for understanding and predicting cell-hydrogel self-organization.
• This AI-driven approach facilitates rational design of moulds for applications in drug testing and regenerative medicine.
• Future work can extend this methodology to scaffolds and 3D bioprinting.