Development of a quantitative prediction algorithm for human cord blood-derived CD34⁺ hematopoietic stem-progenitor...

Area of Science:

• Hematology
• Stem Cell Biology
• Bioinformatics

Background:

• Cord blood stem cell transplantation is a standard treatment for various disorders.
• The dose of CD34+ hematopoietic stem-progenitor cells (HSPCs) is critical for successful outcomes.
• Accurate prediction of HSPC yield is essential for clinical application.

Purpose of the Study:

• To evaluate mathematical models for predicting the proportion of CD34+ cells in cryopreserved cord blood.
• To identify maternal and neonatal parameters that influence HSPC yield.
• To compare the predictive power of parametric and non-parametric algorithms.

Main Methods:

• Analysis of 24 predictor variables from 802 processed cord blood units (2020-2022).
• Development of prediction models using multivariate linear regression (parametric), random forest, and back propagation neural network (non-parametric).
• Evaluation of model performance using root-mean-square deviation and median absolute deviation.

Main Results:

• The back propagation neural network model demonstrated the highest predictive power (56.99%).
• Multivariate linear regression yielded the lowest root-mean-square deviation (0.0982).
• The neural network model showed the highest median absolute deviation (0.0689).

Conclusions:

• Maternal and neonatal parameters can be used to predict the CD34+ cell dose in cord blood products.
• The back propagation neural network offers the most accurate prediction for clinical utilization.
• This predictive model can assist cell banks in selecting optimal cord blood units for transplantation.