High-dimensional Biomarker Identification for Scalable and Interpretable Disease Prediction via Machine Learning...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Complex human diseases are influenced by both omics data and clinical features.
• Identifying key biomarkers and risk factors is crucial for understanding disease mechanisms and advancing precision medicine.
• High-dimensionality and complex associations in omics data pose analytical challenges.

Purpose of the Study:

• To develop a robust framework for biomarker and clinical risk factor identification from high-dimensional omics and clinical data.
• To enhance disease outcome prediction and feature importance identification for complex diseases.
• To provide a computationally efficient tool for researchers in precision medicine.

Main Methods:

• Proposed an ensemble data-driven biomarker identification tool, Hybrid Feature Screening (HFS), to construct a candidate feature set.
• Developed the High-dimensional Feature Importance Test (HiFIT) framework, refining candidate features using a permutation-based importance test.
• Utilized extensive numerical simulations and real-world applications to validate the framework.

Main Results:

• HiFIT demonstrated superior performance in outcome prediction compared to existing methods.
• The framework effectively identified key biomarkers and clinical risk factors.
• Demonstrated computational efficiency in handling high-dimensional data.

Conclusions:

• The HiFIT framework offers a powerful approach for biomarker discovery and risk factor identification in complex diseases.
• HiFIT facilitates advancements in early disease diagnosis and personalized treatment strategies.
• An R package for HiFIT is publicly available to support the research community.