Joint modeling of imaging and genetics.

Area of Science:

• Neuroscience
• Genetics
• Biostatistics

Background:

• Traditional imaging genetics analysis involves separate steps for feature selection and genetic variant identification.
• This sequential approach may not fully capture the complex interplay between imaging phenotypes and genetic factors in disease.

Purpose of the Study:

• To develop a unified Bayesian framework for simultaneous detection of disease-associated genetic variants and intermediate imaging phenotypes.
• To improve the probabilistic assessment of relevance for both genetic and imaging markers.

Main Methods:

• A novel unified Bayesian framework integrating genetic and imaging data.
• Development of an efficient approximate inference algorithm for high-dimensional data.
• Simultaneous modeling of genetic variants and image-based features.

Main Results:

• The proposed method outperforms traditional, two-step approaches on synthetic data.
• Demonstrated the framework's ability to assign probabilistic relevance to genetic and imaging markers.
• Successfully applied the method to real-world Alzheimer's Disease Neuroimaging Initiative (ADNI) data.

Conclusions:

• The unified Bayesian framework offers a more effective approach to imaging genetics analysis.
• Simultaneous modeling enhances the detection of disease-related genetic and imaging markers.
• The method provides a robust tool for investigating the genetic underpinnings of diseases using neuroimaging data.