Identifying disease sensitive and quantitative trait-relevant biomarkers from multidimensional heterogeneous imaging...

Area of Science:

• Neuroscience
• Genetics
• Machine Learning

Background:

• Brain imaging and genotyping advances enable studying genetic and anatomical influences on brain function.
• Traditional methods often overlook complex interactions between neuroimaging, cognition, and disease status.

Purpose of the Study:

• To develop a sparse multimodal multitask learning method for analyzing complex gene-brain-symptom relationships.
• To identify biomarkers integrating genetic and imaging data for predicting cognitive and disease outcomes.

Main Methods:

• Proposed a sparse multimodal multitask learning framework with structured sparsity regularizations.
• Employed a joint classification and regression model for biomarker identification.
• Developed an efficient optimization algorithm with theoretical convergence analysis.

Main Results:

• The method demonstrated improved prediction accuracy for cognitive scores and disease status using Alzheimer's Disease Neuroimaging Initiative data.
• Identified multimodal biomarkers capable of predicting both disease status and cognitive function.
• Elucidated biological pathways from genes to brain structure/function, cognition, and disease.

Conclusions:

• The novel method effectively integrates heterogeneous imaging genetics data to reveal complex biological relationships.
• Identified biomarkers offer insights into the genetic underpinnings of brain function, cognition, and neurological disorders.
• This approach advances the understanding of gene-brain-disease connections and has potential clinical applications.