PGS

PGS applies a penalized regression framework with grid search to identify associations between high-dimensional microRNA (miRNA) expression data and repeated measures for improved phenotype prediction and pathway enrichment.

Key Features:

• Penalized regression framework: Implements a penalized regression approach to model associations in high-dimensional miRNA expression data.
• Grid search parameter tuning: Uses a grid search method to select penalization parameters for the regression model.
• Repeated-measures handling: Analyzes longitudinal/repeated measures miRNA expression data across multiple time points per individual.
• Comparison to site-by-site (SBS) testing: Evaluated against SBS testing and shown to improve analytical outcomes relative to univariate SBS approaches.
• Phenotype prediction performance: Demonstrates smaller phenotype prediction errors compared to SBS testing.
• Pathway enrichment: Produces higher enrichment in phenotype-related biological pathways than SBS testing.
• Simulation validation: Validated by extensive simulations reporting more accurate estimates, increased sensitivity, and comparable specificity.

Scientific Applications:

• Longitudinal miRNA association analysis: Identifies associations between miRNA expression trajectories and phenotypes in repeated-measures studies.
• Phenotype prediction from miRNA profiles: Improves prediction of phenotypic outcomes using high-dimensional miRNA data.
• Pathway-level interpretation: Facilitates enrichment analyses to link miRNA-associated signals to phenotype-related biological pathways.
• Method benchmarking: Serves as a comparative method for evaluating analytic performance against SBS testing using simulations and real datasets.

Methodology:

Implements a penalized regression model combined with a grid search for parameter tuning and is evaluated via comparative analyses against site-by-site (SBS) testing and extensive simulation studies.