An Alzheimers disease related genes identification method based on multiple classifier integration.

Area of Science:

• Genetics
• Neuroscience
• Bioinformatics

Background:

• Alzheimer's disease (AD) is a fatal neurodegenerative disorder with an insidious onset.
• The complete set of Alzheimer's disease-related genes (ADGs) remains incompletely understood.
• Existing datasets contain thousands of genes, with only a fraction currently identified as ADGs.

Purpose of the Study:

• To identify additional, previously undiscovered Alzheimer's disease genes (ADGs).
• To leverage machine learning techniques for enhanced ADG identification.
• To improve the accuracy and sensitivity of ADG detection.

Main Methods:

• Implemented a gene identification method integrating multiple classifiers.
• Utilized a feature selection algorithm to identify the most relevant genetic attributes.
• Developed a two-stage cascading classifier: Relevance Vector Machine followed by a voting ensemble (Support Vector Machine, Random Forest, Extreme Learning Machine).

Main Results:

• Feature selection improved accuracy and reduced computational training time.
• The integrated voting classifier reduced classification errors.
• The system achieved 78.77% accuracy, 83.10% sensitivity, and 74.67% specificity, identifying potentially new ADGs.

Conclusions:

• A novel ADG identification method combining feature selection, cascading classifiers, and majority voting was presented.
• The proposed method significantly enhances accuracy and sensitivity in ADG identification.
• This approach successfully identified potentially new ADGs, advancing AD research.