Deep belief network-based approach for detecting Alzheimer's disease using the multi-omics data.

Area of Science:

• Bioinformatics
• Genomics
• Neuroscience

Background:

• Alzheimer's disease (AD) mechanisms remain poorly understood, with limited identification of genetic risk factors.
• Past research relied heavily on brain imaging, lacking robust genetic analysis techniques.
• Advancements in high-throughput bioinformatics enable focused research into AD-associated genetic risk factors.

Purpose of the Study:

• To develop a Deep Belief Network (DBN) based prediction model for Alzheimer's disease.
• To address High-Dimension Low Sample Size (HDLSS) challenges in genomic and epigenomic data.
• To identify reliable genetic risk factors for Alzheimer's disease using multi-omics data.

Main Methods:

• Utilized DNA Methylation and Gene Expression Microarray Data.
• Implemented a two-layer feature selection approach: identifying differentially expressed/methylated features and combining datasets via Jaccard similarity.
• Applied an ensemble-based feature selection method and a Deep Belief Network for prediction.

Main Results:

• The proposed two-layer feature selection method outperformed Support Vector Machine Recursive Feature Elimination (SVM-RFE) and Correlation-based Feature Selection (CBS).
• The Deep Belief Network model demonstrated superior performance compared to widely used Machine Learning models.
• Multi-omics data integration yielded more promising results than single-omics approaches for AD prediction.

Conclusions:

• The developed feature selection and Deep Belief Network model offer a promising approach for identifying Alzheimer's disease genetic risk factors.
• Multi-omics data integration is effective in enhancing prediction accuracy for complex neurological disorders like AD.
• This study provides a foundation for improved diagnostic and predictive tools for Alzheimer's disease.