Uncovering Key Genes in Neurodegenerative Diseases Through Unsupervised Learning: A Variance-Based LSTM and...

Area of Science:

• Genomics
• Computational Biology
• Neuroscience

Background:

• Neurodegenerative diseases (NDs) like Parkinson's, Alzheimer's, and ALS are complex, with poorly understood genetic underpinnings.
• Identifying significant genes and their interactions is crucial for understanding disease mechanisms and developing targeted therapies.

Purpose of the Study:

• To develop and validate an integrated computational approach for discovering and clustering genes associated with neurodegenerative diseases.
• To identify key biological pathways implicated in Parkinson's disease, Alzheimer's disease, and ALS.

Main Methods:

• Variance-based filtering to select biologically relevant genes.
• Long Short-Term Memory (LSTM) neural networks for capturing gene expression patterns.
• Uniform Manifold Approximation and Projection (UMAP) for dimensionality reduction and K-Means for gene clustering, with Silhouette Score for optimal cluster determination.
• A feedback loop integrating cluster labels into the LSTM model to enhance gene association detection.

Main Results:

• The integrated approach successfully identified and clustered significant genes related to neurodegenerative diseases.
• Enrichment analysis highlighted the involvement of immune regulation and protein signaling pathways in Parkinson's disease, Alzheimer's disease, and ALS.
• Comparison of network architectures demonstrated their impact on clustering performance.

Conclusions:

• The proposed machine learning and clustering framework effectively uncovers meaningful gene associations in neurodegenerative diseases.
• This approach offers valuable insights into the molecular mechanisms underlying these complex conditions.
• The findings pave the way for improved diagnostic and therapeutic strategies for neurodegenerative diseases.