A multivariate regression approach to association analysis of a quantitative trait network

Area of Science:

• Genetics
• Statistical genomics
• Bioinformatics

Background:

• Complex diseases often involve numerous correlated clinical phenotypes, posing challenges for genetic variation identification.
• Traditional association analyses frequently examine phenotypes independently, potentially missing joint genetic influences.

Purpose of the Study:

• To develop a novel statistical framework for identifying genetic variations associated with multiple correlated traits simultaneously.
• To leverage the dependency structure among quantitative traits for improved detection of joint genetic influences.

Main Methods:

• Proposed a graph-guided fused lasso (GFLasso) framework.
• Represented trait dependencies as a network to guide structured regularizations in a multivariate regression model.
• Jointly analyzed all traits within a single statistical model.

Main Results:

• The GFLasso method demonstrated high sensitivity and specificity in detecting genetic markers that jointly influence subgroups of correlated traits.
• Compared to single-marker analysis and other sparse regression methods, GFLasso showed significant advantages in detecting true causal single nucleotide polymorphisms.
• Validation using simulated and asthma datasets confirmed the method's effectiveness in incorporating trait correlation patterns.

Conclusions:

• The graph-guided fused lasso provides a principled and effective approach for multivariate association analysis in complex diseases.
• Incorporating trait correlation structures enhances the power to identify genetic variations influencing multiple related phenotypes.
• The proposed method offers a significant advancement over traditional single-phenotype analyses for complex trait genetics.