LUCIDus

LUCIDus estimates latent clusters and identifies subgroup-specific biomarkers by integratively analyzing genomic, exposure, and metabolomic data to uncover associations and predict risk phenotypes.

Key Features:

• Integrative model: Simultaneously estimates latent clusters and relevant biomarkers/omic effects from combined datasets.
• Latent cluster estimation (EM algorithm): Uses an Expectation-Maximization algorithm to assign latent cluster memberships and estimate model parameters.
• Multi-omics integration: Integrates genomic, exposure, and metabolomic data for joint analysis.
• Feature selection via regularization: Applies regularization methods to identify informative features and reduce noise.
• Subgroup-specific effects: Recapitulates subgroup-specific associations and effects across omics and exposures.
• Predictive modeling: Predicts risk subgroups and phenotypes from integrated omics and exposure data.

Scientific Applications:

• Epidemiologic, clinical, and translational research: Supports analysis of multi-platform omics datasets in epidemiologic, clinical, and translational studies.
• Identifying novel associations: Uncovers associations among genetic markers, environmental exposures, and metabolic profiles.
• Subgroup discovery: Identifies biologically relevant subgroups within populations for mechanistic insight.
• Risk prediction: Enables prediction of risk subgroups and phenotypes to inform stratified analyses.

Methodology:

Implements an integrative model that combines genomic, exposure, and metabolomic data; uses an Expectation-Maximization (EM) algorithm for latent cluster estimation; and applies regularization techniques for feature selection.