Abstract
MOTIVATION: Omics features, often measured by high-throughput technologies, combined with clinical features, significantly impact the understanding of many complex human diseases. Integrating key omics biomarkers with clinical risk factors is essential for elucidating disease mechanisms, advancing early diagnosis, and enhancing precision medicine. However, the high dimensionality and intricate associations between disease outcomes and omics profiles present substantial analytical challenges. RESULTS: We propose a high-dimensional feature importance test (HiFIT) framework to address these challenges. Specifically, we develop an ensemble data-driven biomarker identification tool, Hybrid Feature Screening (HFS), to construct a candidate feature set for downstream machine learning models. The pre-screened candidate features from HFS are further refined using a computationally efficient permutation-based feature importance test employing machine learning methods to flexibly model the potential complex associations between disease outcomes and molecular biomarkers. Through extensive numerical simulation studies and practical applications to microbiome-associated weight changes following bariatric surgery, as well as the examination of gene-expression-associated kidney pan-cancer survival data, we demonstrate HiFIT's superior performance in both outcome prediction and feature importance identification. AVAILABILITY AND IMPLEMENTATION: An R package implementing the HiFIT algorithm is available on GitHub (https://github.com/BZou-lab/HiFIT).