ESPClust

ESPClust identifies covariates that modify the effect size profile in omics association studies through unsupervised analysis of effect size profiles to discover subpopulations with distinct associations.

Key Features:

• Unsupervised identification of effect size modifiers: ESPClust identifies covariates that modify the effect size of associations between sets of omics variables and outcomes without requiring prior knowledge of subpopulation structure.
• Effect size profile analysis: ESPClust analyzes the effect size profile to identify regions in covariate space with different effect size profiles.
• Integration of multiple exposures: ESPClust integrates information from multiple exposures to identify effect size modifiers in datasets too small for traditional univariate stratified analyses.
• Extension of moderation concept: ESPClust extends the concept of moderators to encompass multiple exposures rather than single variables.

Scientific Applications:

• Insulin resistance research: ESPClust uncovered nuanced effect size modifications in insulin resistance data that traditional analyses may overlook.
• COVID-19 symptom manifestation: ESPClust identified effect size modifiers in COVID-19 symptom manifestation data, demonstrating versatility across disease contexts.
• Synthetic data validation: ESPClust was validated on synthetic data to demonstrate its ability to recover known effect size modification patterns.
• Personalized medicine: ESPClust enables discovery of subpopulations with distinct associations, supporting personalized medicine approaches in omics research.

Methodology:

ESPClust performs unsupervised analysis of effect size profiles to identify regions in covariate space with different effect size profiles, extending the concept of moderators to encompass multiple exposures.