Priority-Elastic net for binary disease outcome prediction based on multi-omics data.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• High-dimensional omics data integration is crucial for advancing healthcare predictive models.
• Challenges include data heterogeneity, variable prioritization, information flow assessment, and multicollinearity.

Purpose of the Study:

• To introduce a novel hierarchical regression method for improved multi-omics data integration.
• To address challenges in prioritizing and integrating diverse omics data blocks for enhanced prediction.

Main Methods:

• Proposed the Priority-Elastic net algorithm, a hierarchical regression approach extending Priority-Lasso.
• Incorporated a variable block priority order and sequential Elastic-net fitting.
• Evaluated the Priority-Adaptive Elastic net penalty for comparative analysis.

Main Results:

• The Priority-Elastic net and Priority-Adaptive Elastic net algorithms were tested on a brain tumor dataset (TCGA).
• Data included transcriptomics, proteomics, and clinical information for Lower-grade glioma (LGG) and glioblastoma (GBM).

Conclusions:

• The Priority-Elastic net algorithm demonstrates superior stability and prediction accuracy compared to other methods.
• Offers moderate computational complexity and flexibility for integrating prior knowledge in hierarchical models.
• Provides a significant advancement for personalized medicine through optimized multi-omics data analysis.