T-MTSCCA

T-MTSCCA models associations between longitudinal neuroimaging quantitative traits (QTs) and single nucleotide polymorphisms (SNPs) to identify temporal imaging-genetic patterns underlying brain structure, function, and disorders.

Key Features:

• Longitudinal Data Analysis: Leverages longitudinal neuroimaging datasets across multiple time points to capture dynamic progression of imaging quantitative traits (QTs).
• Multi-Task Sparse Canonical Correlation Analysis (MTSCCA): Implements a temporal multi-task sparse canonical correlation analysis model to jointly relate multiple neuroimaging datasets with single nucleotide polymorphisms (SNPs).
• Identification of Progressive Patterns: Identifies trajectories of progressive imaging-genetic patterns, distinguishing time-consistent and time-dependent SNPs and imaging QTs.
• Efficient Algorithm: Solves the longitudinal optimization problem using an efficient algorithm with proven convergence properties.
• Performance Evaluation: Demonstrated on 408 subjects from the Alzheimer's Disease Neuroimaging Initiative (ADNI), achieving higher canonical correlation coefficients and clearer canonical weight patterns compared to state-of-the-art methods.

Scientific Applications:

• Genetic Basis of Brain Disorders: Links SNPs to longitudinal changes in brain QTs to elucidate genetic contributions to disorders such as Alzheimer's disease.
• Disease Progression Studies: Facilitates analysis of temporal dynamics in genetic influences for progressive neurological diseases.
• Endophenotype Research: Supports identification of imaging endophenotypes by integrating longitudinal imaging QTs with genetic data.

Methodology:

Applies temporal multi-task sparse canonical correlation analysis to integrate longitudinal neuroimaging QTs and SNP data, incorporates relationships within imaging data and SNPs, and solves the resulting optimization via an efficient algorithm with proven convergence.