AI-based approach to dissect the variability of mouse stem cell-derived embryo models.

Area of Science:

• Developmental Biology
• Artificial Intelligence in Biology
• Stem Cell Research

Background:

• Stem cell-derived embryo models offer novel insights into embryogenesis but face challenges in standardization due to developmental variability.
• Reproducible selection of these models is crucial for advancing developmental biology research.

Purpose of the Study:

• To enhance the reproducibility of selecting stem cell-derived embryo models using deep learning.
• To classify mouse post-implantation stem cell-derived embryo-like structures (ETiX-embryos) into normal and abnormal categories using AI.

Main Methods:

• Utilized live imaging and AI-based deep learning models to analyze 900 mouse ETiX-embryos.
• Developed models to classify embryo-like structures at 90 hours post-cell seeding and at the initial cell-seeding stage.
• Conducted perturbation experiments by altering initial cell numbers to assess impact on development.

Main Results:

• The best-performing AI model achieved 88% accuracy at 90 hours and 65% accuracy at the initial seeding stage, forecasting developmental trajectories.
• Identified higher cell counts, larger size, and more compact shape as key features of normally developing ETiX-embryos.
• Perturbation experiments confirmed that increased initial cell numbers improve normal development outcomes.

Conclusions:

• Deep learning significantly improves the selection and standardization of stem cell-derived embryo models.
• The study reveals critical features governing ETiX-embryo self-organization, advancing consistency in the field.
• AI-driven analysis provides a powerful tool for developmental biology research using embryo models.