Machine learning-based scoring models to predict hematopoietic stem cell mobilization in allogeneic donors.

Area of Science:

• Hematology
• Immunology
• Biotechnology

Background:

• Mobilized peripheral blood is the primary source for hematopoietic stem cell transplantation.
• Granulocyte colony-stimulating factor (G-CSF) is the standard mobilization agent, requiring 4-5 days of administration.
• Some donors do not achieve optimal stem cell yields with G-CSF.

Purpose of the Study:

• To compare the efficacy of G-CSF with alternative mobilization regimens.
• To develop machine learning models for predicting stem cell yield.
• To identify donors who may benefit from alternative mobilization strategies.

Main Methods:

• Retrospective review of 1361 related allogeneic donors.
• Comparison of G-CSF with GM-CSF, G-CSF+GM-CSF, GM-CSF+Plerixafor, Plerixafor, and BL-8040.
• Development of machine learning-based scoring models using donor demographic and laboratory data.

Main Results:

• Cytokine-based strategies (G-CSF, GM-CSF) yielded higher CD34+ cell counts after 4-5 days.
• CXCR4 antagonists (Plerixafor, BL-8040) induced rapid, lower cell yields on the same day.
• Machine learning models were developed to predict suboptimal stem cell yields.

Conclusions:

• Alternative mobilization agents offer different kinetics and efficacy compared to G-CSF.
• Machine learning models can predict stem cell yield early in the donor screening process.
• Predictive models can guide the selection of optimal mobilization strategies for enhanced stem cell yields.