Annotating whole genome variants and constructing a multi-classifier based on samples of ADNI

Area of Science:

• Genomics
• Neurodegenerative Diseases
• Computational Biology

Background:

• Alzheimer's disease (AD) is a leading cause of dementia in the elderly, with its complex mechanisms poorly understood.
• Effective treatments and preventative measures for AD are lacking, highlighting the need for further research into its genetic underpinnings.

Purpose of the Study:

• To investigate the relationship between DNA variants and Alzheimer's disease phenotypes using whole genome sequencing data.
• To identify potential susceptibility genes and pathways associated with AD development.
• To develop a robust classification model for distinguishing between AD, mild cognitive impairment (MCI), and cognitive normal (CN) individuals.

Main Methods:

• Annotation of deleterious variants and mapping to nearest protein-coding genes.
• Application of a multi-objective evaluation strategy based on entropy theory for gene ranking.
• Utilizing a multi-classifier XGBoost model for classifying unbalanced datasets (AD, MCI, CN) from the Alzheimer's Disease Neuroimaging Initiative (ADNI).

Main Results:

• The proposed method achieved satisfactory classification performance in distinguishing AD, MCI, and CN samples.
• A significant correlation was identified between Alzheimer's disease and the *RIN3* gene, a known AD susceptibility gene.
• Pathway enrichment analysis highlighted three pathways significantly associated with AD pathogenesis.

Conclusions:

• The developed method demonstrates practical significance in analyzing whole genome sequencing data for AD research.
• The identification of *RIN3* as a key gene provides a potential target for future AD investigations.
• The study contributes to a better understanding of AD's genetic architecture and associated biological pathways.