MUVR2

MUVR2 implements machine learning frameworks that adjust for covariates and assess modeling fitness to improve variable selection and prediction accuracy in omics research.

Key Features:

• Multiple Modeling Approaches: MUVR2 implements elastic net regularized regression, partial least squares, and random forest modeling.
• Covariate Adjustment: MUVR2 allows adjustment for covariates using elastic net modeling.
• Modeling Fitness Assessment: MUVR2 provides a framework for assessing modeling fitness, addressing a previously lacking capability in machine learning.
• Cross-Validation Performance: MUVR2 consistently shows state-of-the-art performance in variable selection while minimizing overfitting compared with other cross-validation strategies.

Scientific Applications:

• Omics Association Studies: MUVR2 examines associations between molecular data and exposures or health conditions.
• Feature Selection: MUVR2 facilitates biological interpretation through machine learning-based feature selection.

Methodology:

MUVR2 uses repeated double cross-validation with elastic net, partial least squares, and random forest modeling approaches. Covariate adjustment is implemented for elastic net but not for partial least squares or random forest.