Modeling multi-stage disease progression and identifying genetic risk factors via a novel collaborative learning...

Area of Science:

• Neuroscience
• Genetics
• Computational Biology

Background:

• Alzheimer's disease (AD) progression is gradual, necessitating staging for diagnosis and treatment.
• Identifying genetic factors influencing AD pathogenesis is crucial.
• Current methods often handle disease staging and genetic variation identification separately.

Purpose of the Study:

• To develop a novel computational method that jointly models Alzheimer's disease progression and identifies genetic risk factors.
• To address the limitation of separate analyses for disease staging and genetic variation identification.

Main Methods:

• Proposed a sparse multi-stage multi-task mixed-effects collaborative longitudinal regression method (MSColoR).
• MSColoR jointly models disease progression as a multi-stage process using longitudinal neuroimaging data.
• Associated fitted disease trajectories with genetic variations at each stage, leveraging genome-wide association study summary statistics.

Main Results:

• MSColoR reduces modeling errors in longitudinal brain imaging genetics studies.
• The method identifies more accurate and significant genetic variations associated with Alzheimer's disease progression.
• Evaluated using synthetic and real longitudinal neuroimaging and genetic data, outperforming existing longitudinal methods.

Conclusions:

• MSColoR offers a powerful computational technique for analyzing longitudinal brain imaging genetics data in Alzheimer's disease research.
• The joint modeling approach enhances understanding of AD pathogenesis and genetic influences.
• Publicly available code facilitates further research and application.