GFlasso

GFlasso applies a graph-guided fused lasso to jointly analyze genotypes and correlated quantitative traits, enabling detection of genetic variants (single nucleotide polymorphisms, SNPs) that influence groups of related phenotypes.

Key Features:

• Graph-Guided Framework: GFlasso represents dependency among quantitative traits as a network and implements the graph-guided fused lasso (GFlasso) for structured regularization.
• Joint Analysis of Correlated Traits: It performs multivariate regression of genotypes on multiple traits jointly to detect single nucleotide polymorphisms (SNPs) that influence subgroups of highly correlated phenotypes.
• Structured Regularization: The method leverages the trait network to impose structured regularization in multivariate regression, encouraging shared genetic effects across connected traits.
• Performance and Validation: The approach has been validated using simulated datasets derived from HapMap consortium data and real-world asthma datasets, with comparative studies demonstrating improved detection of causal SNPs.

Scientific Applications:

• Complex disease syndrome analysis: Studying complex disease syndromes such as asthma to identify genetic variations that jointly influence multiple interrelated clinical phenotypes.
• Genetic architecture and therapeutic insight: Uncovering underlying genetic architecture of multifaceted diseases and informing the development of targeted therapeutic strategies.

Methodology:

Represent trait dependencies as a network and apply the graph-guided fused lasso (GFlasso) as structured regularization in multivariate regression of genotypes on quantitative traits, with validation using simulated datasets derived from the HapMap consortium and real-world asthma datasets in comparative studies.