AlzDiscovery: A computational tool to identify Alzheimer's disease-causing missense mutations using protein structure...

Area of Science:

• Genetics and Bioinformatics
• Neurodegenerative Diseases
• Computational Biology

Background:

• Alzheimer's disease (AD) is a leading cause of dementia, characterized by protein aggregates.
• Missense mutations in AD-related proteins can alter protein function and increase disease risk.
• Existing variant identification methods often overlook the impact of mutations within a protein's 3D structure.

Purpose of the Study:

• To develop a machine learning model for classifying Alzheimer's disease-causing mutations.
• To leverage both sequence and structure-based features for improved variant prediction.
• To provide a valuable resource for AD screening and personalized therapeutic development.

Main Methods:

• Machine learning analysis classifying missense mutations in 21 AD-related proteins.
• Utilized computational tools to assess mutation effects on protein stability.
• Developed a predictive model incorporating sequence and structure features, optimized with sample weight tuning.

Main Results:

• Identified a bias towards destabilizing effects in pathogenic mutations within AD-related proteins.
• Achieved high performance: 0.95 AUC in blind tests and 0.70 AUC in clinical validation, surpassing state-of-the-art methods.
• Feature interpretation highlighted the importance of hydrophobic environments and polar interactions in mutation pathogenicity.

Conclusions:

• The developed model offers superior accuracy in predicting AD-causing missense mutations.
• The study introduces AlzDiscovery, a web server for predicting mutation phenotypes in AD-related proteins.
• Findings support enhanced AD screening and the development of targeted, personalized treatments.