Accelerated estimation and permutation inference for ACE modeling.

Area of Science:

• Neuroimaging Genetics
• Quantitative Genetics
• Brain Imaging

Background:

• Existing neuroimaging genetic tools struggle with brain-wide heritability analysis at scale.
• Current methods for heritability estimation often require computationally intensive iterative optimization.

Purpose of the Study:

• To present a computationally efficient method for brain-wide heritability estimation in twins.
• To introduce Accelerated Permutation Inference for the ACE Model (APACE) as a faster alternative to traditional variance component models.

Main Methods:

• Developed a non-iterative linear regression model transforming twin-pair difference data.
• Replaced iterative variance component models with a direct linear regression approach.
• Combined the method with permutation testing and spatial statistics (e.g., cluster size) for enhanced power.

Main Results:

• The APACE method demonstrated comparable bias, mean squared error, false positive risk, and power to maximum-likelihood methods.
• APACE requires a significantly smaller fraction of computation time compared to existing best practices.
• The method was successfully applied to a functional magnetic resonance imaging (fMRI) working memory dataset.

Conclusions:

• APACE provides a computationally efficient and accurate approach for brain-wide heritability estimation in twins.
• The method significantly reduces computational demands, enabling large-scale neuroimaging genetic studies.
• APACE, particularly with spatial statistics, enhances the power and feasibility of heritability analyses in neuroimaging.