Adaptive Sparse Multi-Block PLS Discriminant Analysis: An Integrative Method for Identifying Key Biomarkers from...

Area of Science:

• Bioinformatics
• Computational Biology
• Systems Biology

Background:

• High-throughput technologies generate vast multi-omics data for disease mechanism exploration.
• Integrating diverse omics data is crucial for understanding host-disease molecular associations.

Purpose of the Study:

• To present adaptive sparse multi-block partial least square discriminant analysis (asmbPLS-DA) for multi-omics data integration.
• To identify relevant features across different omics data types for discriminating multiple disease outcomes.
• To evaluate the performance of asmbPLS-DA in biomarker discovery and disease classification.

Main Methods:

• Developed an extension of asmbPLS, named asmbPLS-DA, for integrative multi-omics analysis.
• Utilized simulation data under various scenarios and a real dataset from The Cancer Genome Atlas (TCGA).
• Compared asmbPLS-DA with existing methods for feature selection and classification performance.

Main Results:

• asmbPLS-DA identified key biomarkers with improved biological relevance compared to other methods.
• The method demonstrated competitive classification performance for disease status and phenotypes, especially when combined with other algorithms.
• An R package, asmbPLS, implementing the method is available on GitHub.

Conclusions:

• asmbPLS-DA is a valuable tool for multi-omics research, excelling in feature selection and classification.
• The integrative approach enhances the understanding of molecular mechanisms underlying diseases.
• The method offers improved biological relevance in biomarker discovery from multi-omics data.