asmbPLS-DA

asmbPLS-DA performs adaptive sparse multi-block partial least squares discriminant analysis to integrate genomics, transcriptomics, proteomics, and metabolomics data for feature selection and disease classification.

Key Features:

• Adaptive sparse multi-block integration: Implements adaptive sparsity within a multi-block partial least squares framework to identify and integrate relevant features across genomics, transcriptomics, proteomics, and metabolomics.
• Discriminant analysis for disease classification: Executes discriminant analysis to distinguish among multiple disease outcome groups using integrated multi-omics profiles.
• Enhanced feature selection validated on simulation and TCGA data: Prioritizes biologically relevant biomarkers and has been evaluated using simulation datasets and TCGA (The Cancer Genome Atlas) real-world data.
• Integration with classification algorithms: Outputs can be combined with classification methods such as linear discriminant analysis and random forest to improve subject classification.

Scientific Applications:

• Biomarker discovery: Identifying statistically and biologically relevant biomarkers across multiple omics layers.
• Disease classification: Stratifying patients into disease outcome groups using integrated molecular profiles.
• Understanding molecular mechanisms: Elucidating interactions across omics layers to investigate disease pathogenesis.

Methodology:

Adaptive sparse multi-block partial least squares discriminant analysis integrates multiple omics blocks and applies adaptive sparsity to prioritize features relevant to disease outcomes; validation used simulation data and TCGA datasets, and results can be paired with linear discriminant analysis or random forest.