Methods for Feature Selection in Down-Selection of Vaccine Regimens Based on Multivariate Immune Response Endpoints

Area of Science:

• Clinical Trials
• Biostatistics
• Immunology

Background:

• Comparing multiple candidate regimens using multiple endpoints is crucial in clinical trials, particularly for HIV vaccine development.
• Existing methods for ranking and selecting vaccine regimens can be challenged by correlated immune response endpoints.

Purpose of the Study:

• To propose novel algorithms for selecting parsimonious sets of immune response endpoints for regimen comparison.
• To enhance the performance of regimen down-selection processes through effective endpoint pre-selection.

Main Methods:

• Developed algorithms for immune response endpoint selection based on importance weights and correlation structures.
• Utilized extensive simulation studies to evaluate the proposed methods.
• Applied the method to a real-world HIV vaccine research case.

Main Results:

• Pre-selection of endpoints significantly improves the performance of subsequent regimen down-selection.
• The proposed algorithms effectively address the challenge of correlated endpoints in multi-endpoint comparisons.
• Demonstrated the practical utility of the method in HIV vaccine research.

Conclusions:

• Novel algorithms for endpoint selection enhance the efficiency and effectiveness of comparing clinical trial regimens.
• The methods offer a valuable approach for dimension reduction in multi-endpoint clinical research.
• Applicable to general clinical research beyond HIV vaccine development for optimizing regimen selection.