High-dimensional Immune Profiles and Machine Learning May Predict Acute Myeloid Leukemia Relapse Early following...

Area of Science:

• Immunology
• Hematology
• Computational Biology

Background:

• Acute myeloid leukemia (AML) relapse post-hematopoietic stem cell transplant (HSCT) significantly impacts patient outcomes.
• Identifying early immune markers for AML relapse is crucial for timely intervention.

Purpose of the Study:

• To identify early immune signatures associated with AML relapse after HSCT.
• To develop a predictive model for AML relapse using immune profiling.

Main Methods:

• Analysis of peripheral blood mononuclear cells (PBMCs) from 58 AML patients undergoing HSCT using high-dimensional flow cytometry.
• Application of Uniform Manifold Approximation and Projection (UMAP) and PhenoGraph clustering for T cell subset analysis.
• Development of a supervised machine learning algorithm (XGBoost with ADASYN) for relapse prediction.

Main Results:

• Distinct changes in CD4+ and CD8+ T cell populations were observed in patients relapsing within one year.
• Increased IL-2, IL-10, and IL-17-producing CD4+ T cells and decreased CD8+ T cell function were noted in relapsing patients.
• A machine learning model achieved 90% accuracy in predicting AML relapse within 30 days post-HSCT.

Conclusions:

• A specific immunological signature can predict AML relapse early after HSCT.
• Machine learning models can leverage immune phenotypic data for accurate relapse prediction.
• These findings offer potential for preemptive strategies against AML recurrence.