Bayesian mixed model inference for genetic association under related samples with brain network phenotype

Area of Science:

• Neuroscience
• Quantitative Genetics
• Statistical Genetics

Background:

• Genetic association studies for brain connectivity are advancing with new imaging and genetics techniques.
• Brain connectivity presents unique challenges due to its network structure and biological complexity.
• Existing models struggle with sample relatedness common in neuroimaging genetics.

Purpose of the Study:

• Propose a Bayesian network-response mixed-effect model for brain connectivity phenotypes.
• Incorporate population structures like pedigrees and unknown sample relatedness.
• Address the limitations of current network-response modeling in imaging genetics.

Main Methods:

• Developed a Bayesian network-response mixed-effect model for network-variate phenotypes.
• Modeled genetic effects using effect network configurations with inter-network sparsity and intra-network shrinkage.
• Employed a Markov chain Monte Carlo (MCMC) algorithm for uncertainty quantification.

Main Results:

• Extensive simulations demonstrated the model's effectiveness.
• Applied the model to Human Connectome Project data on brain structural connectivity.
• Achieved plausible and interpretable results for genetic bases of brain connectivity.

Conclusions:

• The proposed model effectively analyzes genetic influences on brain connectivity, handling complex population structures.
• It provides a robust framework for dissecting genetic contributions to network-variate phenotypes.
• The model offers a general linear mixed-effect regression framework applicable beyond brain connectivity studies.