Identify Huntington's disease associated genes based on restricted Boltzmann machine with RNA-seq data

Area of Science:

• Computational biology
• Genomics
• Machine learning

Background:

• Predicting disease-associated genes aids understanding of molecular mechanisms in disease progression.
• Complex neurodegenerative diseases challenge traditional statistical methods for gene identification.
• Deep learning models show promise in analyzing biological data and gene expression patterns.

Purpose of the Study:

• To propose a deep learning approach for analyzing RNA-seq data in Huntington's disease.
• To develop a novel framework for screening key disease-associated genes using a stacked restricted Boltzmann machine (SRBM).
• To leverage SRBM's hierarchical structure to capture regulatory factor effects.

Main Methods:

• Utilized a stacked restricted Boltzmann machine (SRBM) for RNA-seq data analysis.
• Designed a framework to screen key genes based on SRBM's differential neuron activation and gene energy changes over time.
• Analyzed time-series gene expression datasets for Huntington's disease.

Main Results:

• The SRBM effectively detects important information for differential analysis of time-series gene expression data.
• The novel framework demonstrated improved identification accuracy for disease-associated genes.
• SRBM achieved enhanced prediction precision for top-ranking disease-associated genes compared to existing methods.

Conclusions:

• SRBM is capable of detecting crucial information in time-series gene expression datasets for differential analysis.
• The proposed framework enhances the accuracy of identifying disease-associated genes.
• SRBM-based prediction of top-ranked disease-associated genes shows superior precision over state-of-the-art methods.