A Review of Statistical Methods in Imaging Genetics

Area of Science:

• Biomedical Data Science
• Statistical Genetics
• Neuroimaging Analysis

Background:

• Modern technology enables massive biomedical datasets with multi-modal imaging, genetic, and clinical data.
• Integrating diverse neuroimaging and genomics data is key to understanding genetic influences on brain structure, function, and disease risk.
• Complex brain disorders like Alzheimer's disease require integrated genetic and imaging data for diagnosis and treatment.

Purpose of the Study:

• To review recent statistical techniques for the joint analysis of high-dimensional imaging and genetic data (a big data squared problem).
• To address the computational and theoretical challenges in analyzing large-scale, multi-modal biomedical datasets.
• To encourage statistical researchers to engage with the emerging field of imaging genetics.

Main Methods:

• Review of statistical methods including massive univariate and voxel-wise approaches.
• Examination of reduced rank regression and mixture models for integrated data analysis.
• Discussion of group sparse multi-task regression for complex genetic and imaging data.

Main Results:

• Identified various statistical techniques applicable to imaging genetics.
• Highlighted the challenges posed by high-dimensional data, spatial smoothness, and linkage disequilibrium.
• Provided a foundation for future methodological development in the field.

Conclusions:

• The joint analysis of neuroimaging and genetic data presents significant challenges but offers transformative potential for understanding brain disorders.
• Recent statistical developments offer promising avenues for tackling the big data squared problem in imaging genetics.
• Further statistical research is needed to fully leverage integrated neuroimaging and genetic data for biomedical discovery.