Identification of Alzheimer associated differentially expressed gene through microarray data and transfer...

Area of Science:

• Neuroscience
• Genetics
• Computational Biology

Background:

• Mental stress and neurodegeneration, including tau hyperphosphorylation and amyloid-beta production, are major factors in late-onset Alzheimer's disease (AD).
• Animal stress models can mimic AD-related neurodegenerative processes, providing a platform for research.

Purpose of the Study:

• To identify differentially expressed genes (DEGs) associated with AD progression in rat models.
• To develop a predictive image analysis model using deep learning to correlate gene expression with histopathological findings in AD.

Main Methods:

• Downloaded and analyzed gene expression data (GSE72062, GSE85162, GSE143951, GSE85238) from NCBI GEO archive to identify DEGs.
• Constructed functional enrichment, pathway, gene signal, protein-protein interaction, and micro-RNA interaction networks for DEGs.
• Utilized a Convolutional Neural Network (CNN) model (VGG16 with transfer learning) for image analysis of rat brain histopathology slides to predict DEGs.

Main Results:

• Identified 10 potential gene targets (ARHGAP32, GNA11, NR5A1, GNAT3, FOSL1, HELZ2, NMUR2, BDKRB1, RPL3L, RPL39L) for controlling neurodegeneration in sporadic AD.
• Developed a predictive model with 89% training accuracy and 61% test accuracy, achieving a minimum loss of 2.480%.
• Discovered a functional relationship between ARHGAP32 and known AD-related genes BCL2 and MMP9.

Conclusions:

• The study successfully integrates microarray data analysis with deep learning-based image analysis for preclinical AD research.
• The developed model can predict differentially expressed genes from histopathology slides, offering a novel approach to understanding molecular changes in AD.
• This methodology enhances traditional preclinical research by enabling image analysis to determine the molecular makeup of biological samples.