Machine Learning and Metabolomics Predict Mesenchymal Stem Cell Osteogenic Differentiation in 2D and 3D Cultures

Area of Science:

• Biotechnology and Regenerative Medicine
• Metabolomics and Machine Learning Applications
• Stem Cell Differentiation Analysis

Background:

• Stem cells are crucial for artificial bone grafts, but differentiation variability poses challenges.
• Ensuring consistent quality of stem cell products is vital for predictable patient outcomes.
• Omics analyses offer tools to evaluate stem cell attributes during biomanufacturing.

Purpose of the Study:

• To benchmark osteogenic differentiation quality using metabolomics and machine learning.
• To develop a robust evaluation tool for stem cell-based bone graft products.
• To assess the utility of these methods in both 2D and 3D cell cultures.

Main Methods:

• Gas chromatography-mass spectrometry (GC-MS) for metabolomics profiling.
• Development and training of an XGboost machine learning model using 11 key metabolites.
• Benchmarking the model's performance on independent 2D and 3D cell culture samples.

Main Results:

• The XGboost model accurately distinguished differentiated from undifferentiated umbilical cord blood mesenchymal stem cells (UCB MSCs).
• The model demonstrated high performance (AUC 82.6%) in capturing osteogenesis in 3D UCB MSC cultures.
• The model showed poor performance (AUC 56.2%) for Wharton's Jelly MSCs, reflecting their known osteogenic underperformance.
• Mineralization strongly correlated with fumarate, glycerol, and myo-inositol levels.

Conclusions:

• Metabolomics combined with machine learning provides a reliable method for evaluating stem cell differentiation quality.
• This approach can serve as a potency assay for Advanced Therapy Medicinal Products.
• The findings support the development of standardized quality control measures for stem cell therapies.