Tree-Based Methods for Discovery of Association between Flow Cytometry Data and Clinical Endpoints

Area of Science:

• Immunology
• Computational Biology
• Biostatistics

Background:

• Flow cytometry is crucial for monitoring immune status in HIV-1 patients.
• CD4 T-cell recovery is a key indicator of antiretroviral therapy (ART) effectiveness.
• Predicting CD4 T-cell recovery aids in personalized HIV treatment strategies.

Purpose of the Study:

• To apply and compare three tree-based algorithms (CART, RFs, LR) for analyzing flow cytometry data.
• To identify key predictors of CD4 T-cell recovery in HIV-1 infected individuals with baseline CD4 counts between 200-350 cells/μL.
• To compare tree-based methods with traditional contingency table analysis.

Main Methods:

• Application of Classification and Regression Trees (CART), Random Forests (RFs), and Logic Regression (LR) to flow cytometry data.
• Analysis focused on predicting CD4 T-cell count recovery post-ART initiation.
• Comparison of findings with standard contingency table analysis.

Main Results:

• Tree-based methods, particularly CART and LR, identified combinations of immune markers predictive of CD4 T-cell recovery.
• Baseline CD3-DR-CD56+CD16+ was consistently identified as an important predictor.
• Immune activation states, especially CD8 T-cell activation, emerged as strong predictors when analyzed in combination.

Conclusions:

• Tree-based algorithms offer advanced analytical capabilities for complex flow cytometry datasets.
• These methods can uncover novel associations and combinations of immune markers not apparent through univariate analysis.
• Tree-based approaches enhance the understanding of factors influencing CD4 T-cell recovery in HIV-1 patients.