Shared components of heritability across genetically correlated traits

Area of Science:

• Genetics
• Biostatistics
• Computational Biology

Background:

• Most genetic variants influencing disease risk affect multiple traits (pleiotropy).
• Pleiotropic associations can reveal shared biological mechanisms underlying genetically correlated traits.
• Existing methods struggle to accurately identify these shared genetic components.

Purpose of the Study:

• To develop and validate a novel method, pleiotropic decomposition regression (PDR), for identifying shared genetic components and their causal variants.
• To apply PDR to complex diseases like coronary artery disease (CAD), asthma, and type II diabetes (T2D) to uncover biologically interpretable mechanisms.
• To improve the estimation of genetic effect sizes by leveraging shared heritability.

Main Methods:

• Developed pleiotropic decomposition regression (PDR) to identify shared genetic components and underlying variants.
• Validated PDR using simulated data, comparing its performance against existing methods.
• Applied PDR to three clusters of genetically correlated traits associated with CAD, asthma, and T2D.
• Assigned genetic variants to identified components and calculated posterior-mean effect sizes.
• Performed out-of-sample validation to assess the predictive accuracy of PDR-estimated effect sizes.

Main Results:

• PDR successfully identified biologically interpretable components for CAD, asthma, and T2D.
• For CAD, PDR revealed components linked to BMI, hypertension, and cholesterol, clarifying their interrelationships.
• PDR significantly improved the correlation (r²) between true and estimated genetic effect sizes in out-of-sample validation.
• Estimated effect size correlations improved by 94% for asthma and 70% for T2D, with a predicted 300% improvement for CAD.

Conclusions:

• PDR is an effective method for dissecting shared heritability and identifying pleiotropic mechanisms.
• The method enhances the accuracy and power of genetic association studies by pooling statistical information across traits.
• PDR provides a robust framework for understanding the genetic architecture of complex diseases and their risk factors.