Predicting in vitro human mesenchymal stromal cell expansion based on individual donor characteristics using machine...

Area of Science:

• Biotechnology and Regenerative Medicine
• Computational Biology and Bioinformatics
• Cell Biology and Tissue Engineering

Background:

• Human mesenchymal stromal cells (hMSCs) are crucial for cell-based therapies, requiring in vitro expansion to achieve clinical quantities.
• Cell expansion is influenced by donor characteristics (age, gender) and culture conditions, necessitating predictive models.
• Computational modeling, particularly machine learning, can elucidate the complex relationships between donor attributes and cell growth dynamics.

Purpose of the Study:

• To develop a predictive computational model for in vitro hMSC expansion dynamics.
• To assess the impact of individual donor characteristics on cell expansion.
• To improve the accuracy of predicting population doubling time (PDT) during cell culture.

Main Methods:

• Utilized a dataset of hMSCs from 174 donors (age 3-64, passages 2-27).
• Applied the Random Forests (RF) machine learning technique to model cell expansion.
• Predicted population doubling time (PDT) for each passage using donor-specific characteristics.

Main Results:

• The RF model demonstrated significantly lower mean absolute error in PDT prediction compared to theoretical estimates or historical data.
• Statistical analysis revealed significant differences in population doubling (PD) and PDT across age categories.
• The youngest donor group (under 10 years) exhibited distinct PD and PDT compared to older groups.

Conclusions:

• A predictive computational model was developed to describe in vitro hMSC expansion dynamics based on donor characteristics.
• This approach shows potential for automating cell expansion culture processes.
• Understanding donor-specific factors is key to optimizing hMSC expansion for therapeutic applications.