Identification of Genetic Risk Factors Based on Disease Progression Derived From Longitudinal Brain Imaging...

Area of Science:

• Neuroimaging Genetics
• Computational Biology
• Neurodegenerative Disorders Research

Background:

• Neurodegenerative disorders progress over time, making cross-sectional studies insufficient for identifying genetic risk factors.
• Existing longitudinal imaging genetic methods often overlook disease progression trajectories, which may serve as more stable disease signatures.
• Accurate identification of genetic risk factors is crucial for understanding and potentially treating these progressive conditions.

Purpose of the Study:

• To propose a novel computational method, SMMLING, for robustly identifying genetic risk factors in longitudinal neuroimaging data.
• To jointly model disease progression and identify genetic associations, leveraging the stability of progression trajectories.
• To enhance the accuracy and relevance of identified genetic risk factors compared to existing longitudinal approaches.

Main Methods:

• Developed a sparse multi-task mixed-effects longitudinal imaging genetic method (SMMLING).
• Modeled disease progression using longitudinal imaging phenotypes and associated fitted trajectories with genetic variations.
• Employed l2,1-norm and fused group lasso (FGL) penalties for individual and group-level loci identification.
• Utilized an efficient optimization algorithm guaranteeing global optimum convergence.

Main Results:

• SMMLING demonstrated decreased modeling error compared to existing longitudinal methods on synthetic and real data.
• The method identified more accurate and relevant genetic factors, with many risk loci missed by comparison methods.
• Baseline status and changing rates (intercept and slope) of progression trajectories proved effective for discovering genetic loci.

Conclusions:

• SMMLING offers a superior and stable approach for identifying genetic risk factors in longitudinal neuroimaging studies.
• The joint modeling of disease progression and genetic variations enhances the discovery of relevant genetic loci.
• SMMLING represents a promising computational tool for advancing neuroimaging genetics research in neurodegenerative disorders.