Deep learning for predicting stem cell efficiency for use in beta cell differentiation

Area of Science:

• Biotechnology
• Regenerative Medicine
• Computational Biology

Background:

• Cell therapy holds promise for diabetes treatment, but stem cell production is a bottleneck.
• Variability in stem cell differentiation and difficulty in early identification of efficient clones hinder progress.
• Current methods require expert assessment, which is time-consuming and prone to error.

Purpose of the Study:

• To develop an image-based deep learning model for early prediction of stem cell clone efficiency.
• To guide the selection of highly efficient stem cell clones for improved diabetes cell therapy.
• To reduce the cost and increase the success rate of pancreatic [Formula: see text] -cell production.

Main Methods:

• Application of various deep learning models to phase-contrast images of stem cell clones.
• Classification of stem cell clones based on learned morphological differences.
• Utilizing layer-wise relevance propagation and Fourier-based frequency analysis for feature interpretation.
• Employing an EfficientNet-V2-S model for prediction.

Main Results:

• Achieved 96.7% accuracy in predicting clone efficiency at 53 hours.
• Demonstrated the model's ability to differentiate between high- and low-efficiency stem cell clones.
• Identified cell population structure as a key predictive feature.

Conclusions:

• Deep learning combined with label-free imaging offers a highly predictive method for stem cell clone selection.
• This approach can significantly optimize the production of pancreatic [Formula: see text] -cells for diabetes therapy.
• The study serves as a proof-of-concept for AI-driven enhancement of cell therapy manufacturing.