Page 63 - BH-2-4

Brain & Heart Alzheimer’s disease: Gene and protein network analysis

study on AD and focuses on the molecular mechanisms clusterProfiler package, and the overrepresentation of these
underlying NFT formation, a hallmark of AD pathology. terms was tested with a hypergeometric distribution model.
In this dataset, laser capture microdissection was utilized The analysis was conducted using the human genome as
to select 1,000 neurons with NFTs and 1,000 normal a reference to clarify the specificity of the functions and
neurons from the entorhinal cortex of 10 patients with pathways enriched in AD. Following the enrichment
mid-stage AD (https://www.ncbi.nlm.nih.gov/geo/query/ process, the results were visualized with enrichplot and
acc.cgi?acc=GSE4757). 18-20 Each patient’s contribution ggplot2 packages. Specifically, enrichplot was used to
to the neuron pool was carefully balanced to ensure generate bubble plots that comprehensively displayed the
representativeness. Pooling was performed by combining enriched GO terms, incorporating factors such as gene
the extracted RNA from the respective neuron types across ratio and significance levels, while ggplot2 was utilized to
all patients before conducting the microarray analysis. create bar plots that concisely display the most significant
Here, the term “normal” refers to neurons without visible GO terms based on their p-values. Supplementary File lists
neurofibrillary tangles when examined under a microscope. the genes that were found to be significantly upregulated
Although these neurons are from a diseased environment, in the AD samples compared to control samples. Each
they do not exhibit the specific pathological hallmark entry includes the gene’s identifier, common name, log
(NFTs) and thus serve as a comparative baseline within the fold change quantifying the level of upregulation, and
scope of our study. Mid-stage AD aligns with the moderate the p-value indicating the statistical significance of this
stage of the disease, characterized by more pronounced change. The genes listed here meet the criteria of absolute
deterioration in cognitive functions, significantly impairing log fold change (|log FC|) ≥1 and p < 0.05, highlighting
patients’ ability to perform daily activities independently their potential relevance in AD pathology.
but not yet encompassing the severe end-stage symptoms.
GSE4757 was downloaded from the GEO database with the 2.4. PPI network construction
GEOquery package or analogous tools. To explore the molecular interplay underlying AD, the PPI
network was constructed and analyzed with an approach
2.2. DEG analysis that integrated the STRING database and Cytoscape

To screen AD-related DEGs, bioinformatics analyses were software. This integration provided a platform for the
conducted using the R package “limma,” which is integral to visualization of molecular interaction networks. The use of
the Bioconductor project. The microarray dataset GSE4757 the STRING database ensured that only interactions with
acquired from the GEO database 18,20 was subjected to substantial evidence were included in our analysis, as it
17
data preprocessing, normalization, and quality control. is instrumental in obtaining high-confidence interactions
Subsequent to preprocessing, the ‘limma’ package was based on a predefined threshold. After obtaining the PPI
used to construct a linear model for each gene, and gene network using STRING, the data were imported into
expression was compared between AD and control samples Cytoscape, a versatile tool for analyzing and visualizing
to identify genes with significantly altered expression. The complex networks. The Molecular Complex Detection
empirical Bayes method in ‘limma’ was then applied to (MCODE) plugin in Cytoscape, which operates by scoring
minimize the standard errors of the estimated log-fold network nodes based on local neighborhood density and
changes, thereby enhancing the reliability of the inferences recursively expands clusters based on pre-set parameters,
drawn from the dataset. DEGs 21-23 were selected with was applied to identify densely connected regions. These
stringent criteria: absolute log-fold change (|log FC|) ≥1 regions are indicative of molecular complexes or significant
and p < 0.05, thus ensuring that only the most statistically biological modules in the large network.
significant genes were included.
2.5. Hub gene identification
2.3. GO enrichment analysis The CytoHubba plugin in Cytoscape is a versatile tool
26
To determine AD-related gene expression profiles, for identifying hub genes and significant molecular
the GO enrichment analysis was performed with the interactions, offering a variety of ranking algorithms
R clusterProfiler package, a powerful tool within the to assess the centrality of nodes within the network.
Bioconductor project designed for comparing and CytoHubba was utilized to analyze the intricate interactions
visualizing biological themes among gene clusters, to among enriched genes. Considering factors such as degree,
decipher the biological functions and pathways significantly betweenness, or closeness centrality, an appropriate
associated with the disease. 24,25 The enrichment analysis algorithm was selected to determine hub genes in AD.
was carried out after the identification of DEGs from the Subsequently, the interconnectivity among these hub genes
dataset. The DEGs were mapped to GO terms with the was analyzed and visualized with Cytoscape.

Volume 2 Issue 4 (2024) 3 doi: 10.36922/bh.2906