Analytical Strategy to Prioritize Alzheimer's Disease Candidate Genes in Gene Regulatory Networks Using Public Expression Data

DeepGAMI: deep biologically guided auxiliary learning for multimodal integration and imputation to improve genotype-phenotype prediction

BACKGROUND

Genotypes are strongly associated with disease phenotypes, particularly in brain disorders. However, the molecular and cellular mechanisms behind this association remain elusive. With emerging multimodal data for these mechanisms, machine learning methods can be applied for phenotype prediction at different scales, but due to the black-box nature of machine learning, integrating these modalities and interpreting biological mechanisms can be challenging. Additionally, the partial availability of these multimodal data presents a challenge in developing these predictive models.

METHOD

To address these challenges, we developed DeepGAMI, an interpretable neural network model to improve genotype-phenotype prediction from multimodal data. DeepGAMI leverages functional genomic information, such as eQTLs and gene regulation, to guide neural network connections. Additionally, it includes an auxiliary learning layer for cross-modal imputation allowing the imputation of latent features of missing modalities and thus predicting phenotypes from a single modality. Finally, DeepGAMI uses integrated gradient to prioritize multimodal features for various phenotypes.

RESULTS

We applied DeepGAMI to several multimodal datasets including genotype and bulk and cell-type gene expression data in brain diseases, and gene expression and electrophysiology data of mouse neuronal cells. Using cross-validation and independent validation, DeepGAMI outperformed existing methods for classifying disease types, and cellular and clinical phenotypes, even using single modalities (e.g., AUC score of 0.79 for Schizophrenia and 0.73 for cognitive impairment in Alzheimer's disease).

CONCLUSION

We demonstrated that DeepGAMI improves phenotype prediction and prioritizes phenotypic features and networks in multiple multimodal datasets in complex brains and brain diseases. Also, it prioritized disease-associated variants, genes, and regulatory networks linked to different phenotypes, providing novel insights into the interpretation of gene regulatory mechanisms. DeepGAMI is open-source and available for general use.

Transcriptomic Analyses of Exercise Training in Alzheimer's Disease Cerebral Cortex

BACKGROUND

Research reported exercise could reduce Alzheimer's disease (AD) symptoms in human and animals. However, the molecular mechanism of exercise training via transcriptomic analysis was unclear especially in AD in the cortex area.

OBJECTIVE

Investigate potential significant pathways in the cortex area that were affected by exercise during AD.

METHODS

RNA-seq analysis, differential expressed genes, functional enrichment analysis, and GSOAP clustering analysis were performed in the isolated cerebral cortex from eight 3xTg AD mice (12 weeks old) randomly and equally divided into control (AD) and exercise training (AD-EX) group. Swimming exercise training in AD-EX group was conducted 30 min/day for 1 month.

RESULTS

There were 412 genes significant differentially expressed in AD-EX group compared to AD group. Top 10 upregulated genes in AD-EX group against AD group mostly correlated with neuroinflammation, while top 10 downregulated genes mostly had connection with vascularization, membrane transport, learning memory, and chemokine signal. Pathway analysis revealed the upregulated interferon alpha beta signaling in AD-EX had association with cytokines delivery in microglia cells compared to AD and top 10 upregulated genes involved in interferon alpha beta were Usp18, Isg15, Mx1, Mx2, Stat1, Oas1a, and Irf9; The downregulated extracellular matrix organization in AD-EX had correlation with Aβ and neuron cells interaction and Vtn was one of the top 10 downregulated genes involved in this pathway.

CONCLUSION

Exercise training influenced 3xTg mice cortex through interferon alpha beta signaling upregulation and extracellular matrix organization downregulation based on transcriptomics analysis.

Biomarker Candidates for Alzheimer’s Disease Unraveled through In Silico Differential Gene Expression Analysis

Alzheimer’s disease (AD) is neurodegeneration that accounts for 60–70% of dementia cases. Symptoms begin with mild memory difficulties and evolve towards cognitive impairment. The underlying risk factors remain primarily unclear for this heterogeneous disorder. Bioinformatics is a relevant research tool that allows for identifying several pathways related to AD. Open-access databases of RNA microarrays from the peripheral blood and brain of AD patients were analyzed after background correction and data normalization; the Limma package was used for differential expression analysis (DEA) through statistical R programming language. Data were corrected with the Benjamini and Hochberg approach, and genes with p-values equal to or less than 0.05 were considered to be significant. The direction of the change in gene expression was determined by its variation in the log2-fold change between healthy controls and patients. We performed the functional enrichment analysis of GO using goana and topGO-Limma. The functional enrichment analysis of DEGs showed upregulated (UR) pathways: behavior, nervous systems process, postsynapses, enzyme binding; downregulated (DR) were cellular component organization, RNA metabolic process, and signal transduction. Lastly, the intersection of DEGs in the three databases showed eight shared genes between brain and blood, with potential use as AD biomarkers for blood tests.

Role of adaptin protein complexes in intracellular trafficking and their impact on diseases

Adaptin proteins (APs) play a crucial role in intracellular cell trafficking. The 'classical' role of APs is carried out by AP1‒3, which bind to clathrin, cargo, and accessory proteins. Accordingly, AP1-3 are crucial for both vesicle formation and sorting. All APs consist of four subunits that are indispensable for their functions. In fact, based on studies using cells, model organism knockdown/knock-out, and human variants, each subunit plays crucial roles and contributes to the specificity of each AP. These studies also revealed that the sorting and intracellular trafficking function of AP can exert varying effects on pathology by controlling features such as cell development, signal transduction related to the apoptosis and proliferation pathways in cancer cells, organelle integrity, receptor presentation, and viral infection. Although the roles and functions of AP1‒3 are relatively well studied, the functions of the less abundant and more recently identified APs, AP4 and AP5, are still to be investigated. Further studies on these APs may enable a better understanding and targeting of specific diseases.APs known or suggested locations and functions.

Proteomic profile of human stem cells from dental pulp and periodontal ligament

Background The study of molecular profiling of dental pulp stem cells (DPSCs) and periodontal ligament stem cells (PDLSCs) contributes to understanding the high proliferation ability and multi-lineage differentiation potential. Objectives The aim of the study was to compare the protein abundance and specific markers of DPSCs and PDLSCs by protein profiles. Material and methods The DPSCs and PDLSCs extracted from the same tooth were lysed with 3 biological replicates and the protein was collected. Two-dimensional electrophoresis technology and TMT proteomics were used to separate and identify proteins. The data are available via ProteomeXchange with identifier PXD021997. The RT-qPCR detection of mRNA expression revealed a special marker for distinguishing two kinds of dental stem cells. Results Compared with PDLSCs, 962 differential proteins (DAPs) were up-regulated, and 127 were down-regulated in DPSCs. In the up-regulated DAPs, two high-scoring sub-networks were detected for neural-related molecules, which encode cell vesicle transport and mitochondrial energy transfer to regulate cell proliferation and secretion factors. A large number of cell adhesion molecules were distinguished among the highly expressed molecules of PDLSCs, supporting that stem cells provide cell attachment functions. It was interpreted ENPL, HS90A and HS90B were highly expressed in DPSCs, while CKB was highly abundant in PDLSCs. Another cell group confirmed that these molecules can be used as special biomarkers to identify and distinguish between DPSCs and PDLSCs. Conclusions This study can promote the basic research and clinical application of dental stem cells. Significance The high-throughput protein profiles were tested by combining two-dimensional gel proteomics and TMT-based proteomics. The proteomics of DPSCs and PDLSCs without individual difference demonstrated an accurate and comprehensive molecular expression profiles and interpretation of neural application potential, this study promotes the basic research of dental stem cells and clinical application.

Behavioral sensitization induced by methamphetamine causes differential alterations in gene expression and histone acetylation of the prefrontal cortex in rats

Background

Methamphetamine (METH) is one of the most widely abused illicit substances worldwide; unfortunately, its addiction mechanism remains unclear. Based on accumulating evidence, changes in gene expression and chromatin modifications might be related to the persistent effects of METH on the brain. In the present study, we took advantage of METH-induced behavioral sensitization as an animal model that reflects some aspects of drug addiction and examined the changes in gene expression and histone acetylation in the prefrontal cortex (PFC) of adult rats.

Methods

We conducted mRNA microarray and chromatin immunoprecipitation (ChIP) coupled to DNA microarray (ChIP-chip) analyses to screen and identify changes in transcript levels and histone acetylation patterns. Functional enrichment analyses, including Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses, were performed to analyze the differentially expressed genes. We then further identified alterations in ANP32A (acidic leucine-rich nuclear phosphoprotein-32A) and POU3F2 (POU domain, class 3, transcription factor 2) using qPCR and ChIP-PCR assays.

Results

In the rat model of METH-induced behavioral sensitization, METH challenge caused 275 differentially expressed genes and a number of hyperacetylated genes (821 genes with H3 acetylation and 10 genes with H4 acetylation). Based on mRNA microarray and GO and KEGG enrichment analyses, 24 genes may be involved in METH-induced behavioral sensitization, and 7 genes were confirmed using qPCR. We further examined the alterations in the levels of the ANP32A and POU3F2 transcripts and histone acetylation at different periods of METH-induced behavioral sensitization. H4 hyperacetylation contributed to the increased levels of ANP32A mRNA and H3/H4 hyperacetylation contributed to the increased levels of POU3F2 mRNA induced by METH challenge-induced behavioral sensitization, but not by acute METH exposure.

Conclusions

The present results revealed alterations in transcription and histone acetylation in the rat PFC by METH exposure and provided evidence that modifications of histone acetylation contributed to the alterations in gene expression caused by METH-induced behavioral sensitization.

Identifying the function of methylated genes in Alzheimer’s disease to determine epigenetic signatures: a comprehensive bioinformatics analysis

Gene methylation is one means of controlling tissue gene expression, but it is unknown what pathways influencing Alzheimer’s disease (AD) are controlled this way. We compared normal and AD brain tissue data for gene expression (mRNAs) and gene methylation profiling. We identified methylated differentially expressed genes (MDEGs). Protein-protein interaction (PPI) of the MDEGs showed 18 hypermethylated low-expressed genes (Hyper-LGs) involved in cell signaling and metabolism; also 10 hypomethylated highly expressed (Hypo-HGs) were involved in regulation of transcription and development. Molecular pathways enriched in Hyper-LGs included neuroactive ligand-receptor interaction pathways. Hypo-HGs were notably enriched in pathways including hippo signaling. PPI analysis also identified both Hyper-LGs and Hypo-HGs, as hub proteins. Our analysis of AD datasets identified Hyper-LGs, Hypo-HGs, and transcription factors linked to these genes. These pathways, which may participate in Alzheimer’s disease development, may be affected by treatments that influence gene methylation patterns.

A Systematic Bioinformatics Workflow With Meta-Analytics Identified Potential Pathogenic Factors of Alzheimer's Disease

Potential pathogenic factors, other than well-known APP, APOE4, and PSEN, can be further identified from transcriptomics studies of differentially expressed genes (DEGs) that are specific for Alzheimer’s disease (AD), but findings are often inconsistent or even contradictory. Evidence corroboration by combining meta-analysis and bioinformatics methods may help to resolve existing inconsistencies and contradictions. This study aimed to demonstrate a systematic workflow for evidence synthesis of transcriptomic studies using both meta-analysis and bioinformatics methods to identify potential pathogenic factors. Transcriptomic data were assessed from GEO and ArrayExpress after systematic searches. The DEGs and their dysregulation states from both DNA microarray and RNA sequencing datasets were analyzed and corroborated by meta-analysis. Statistically significant DEGs were used for enrichment analysis based on KEGG and protein–protein interaction network (PPIN) analysis based on STRING. AD-specific modules were further determined by the DIAMOnD algorithm, which identifies significant connectivity patterns between specific disease-associated proteins and non-specific proteins. Within AD-specific modules, the nodes of highest degrees (>95th percentile) were considered as potential pathogenic factors. After systematic searches of 225 datasets, extensive meta-analyses among 25 datasets (21 DNA microarray datasets and 4 RNA sequencing datasets) identified 9,298 DEGs. The dysregulated genes and pathways in AD were associated with impaired amyloid-β (Aβ) clearance. From the AD-specific module, Fyn, and EGFR were the most statistically significant and biologically relevant. This meta-analytical study suggested that the reduced Aβ clearance in AD pathogenesis was associated with the genes encoding Fyn and EGFR, which were key receptors in Aβ downstream signaling.

coMethDMR: accurate identification of co-methylated and differentially methylated regions in epigenome-wide association studies with continuous phenotypes

Recent technology has made it possible to measure DNA methylation profiles in a cost-effective and comprehensive genome-wide manner using array-based technology for epigenome-wide association studies. However, identifying differentially methylated regions (DMRs) remains a challenging task because of the complexities in DNA methylation data. Supervised methods typically focus on the regions that contain consecutive highly significantly differentially methylated CpGs in the genome, but may lack power for detecting small but consistent changes when few CpGs pass stringent significance threshold after multiple comparison. Unsupervised methods group CpGs based on genomic annotations first and then test them against phenotype, but may lack specificity because the regional boundaries of methylation are often not well defined. We present coMethDMR, a flexible, powerful, and accurate tool for identifying DMRs. Instead of testing all CpGs within a genomic region, coMethDMR carries out an additional step that selects co-methylated sub-regions first. Next, coMethDMR tests association between methylation levels within the sub-region and phenotype via a random coefficient mixed effects model that models both variations between CpG sites within the region and differential methylation simultaneously. coMethDMR offers well-controlled Type I error rate, improved specificity, focused testing of targeted genomic regions, and is available as an open-source R package.

Gene Regulatory Networks in Peripheral Mononuclear Cells Reveals Critical Regulatory Modules and Regulators of Multiple Sclerosis

Multiple sclerosis (MS) is a complex, demyelinating disease with the involvement of autoimmunity and neurodegeneration. Increasing efforts have been made towards identifying the diagnostic markers to differentiate the classes of MS from other similar neurological conditions. Using a systems biology approach, we constructed four types of gene regulatory networks (GRNs) involved in peripheral blood mononuclear cells (PBMCs). The regulatory strength of each GRN across primary progressive MS (PPMS), relapsing-remitting MS (RRMS), secondary progressive MS (SPMS), and control were evaluated by an integrity algorithm. Among the constructed GRNs (referred as TF_gene_miRNA), POU3F2_CDK6_hsa-miR-590-3p, MEIS1_CASC3_hsa-miR-1261, STAT3_OGG1_hsa-miR-298, and TCF4_FMR1_hsa-miR-301b were top-ranked and differentially regulated in all classes of MS compared to control. These GRNs showed potential involvement in regulating various molecular pathways such as interleukin, integrin, glypican, sphingosine phosphate, androgen, and Wnt signaling pathways. For validation, the qPCR analysis of the GRN components (TFs, gene, and miRNAs) in PBMCs of healthy controls (n = 30), RRMS (n = 14), PPMS (n = 13) and SPMS (n = 12) were carried out. Real-time expression analysis of GRNs showed a similar regulatory pattern as derived from our systems biology approach. Also, our study provided several novel GRNs that regulate unique and common molecular mechanisms between MS conditions. Hence, these regulatory components of GRNs will help to understand the disease mechanism across MS classes and further insight may though light towards diagnosis.

Development and use of a Cytoscape app for GRNCOP2

BACKGROUND AND OBJECTIVE

Gene regulatory networks (GRNs) are essential for understanding most molecular processes. In this context, the so-called model-free approaches have an advantage modeling the complex topologies behind these dynamic molecular networks, since most GRNs are difficult to map correctly by any other mathematical model. Abstract model-free approaches, also known as rule-based extraction methods, offer valuable benefits when performing data-driven analysis; such as requiring the least amount of data and simplifying the inference of large models at a faster analysis speed. In particular, GRNCOP2 is a combinatorial optimization method with an adaptive criterion for the discretization of gene expression data and high performance, in contrast to other rule-based extraction methods for discovering GRNs. However, the analysis of the large relational structures of the networks inferred by GRNCOP2 requires the support of effective tools for interactive network visualization and topological analysis of the extracted associations. This need motivated the possibility of integrating GRNCOP2 in the Cytoscape ecosystem in order to benefit from Cytoscapes core functionality, as well as all the other apps in its ecosystem.

METHODS

In this paper, we introduce the implementation of a GRNCOP2 Cytoscape app. This incorporation to Cytoscape platform includes new functionality for GRN visualizations, dynamic user-interaction and integration with other apps for topological analysis of the networks.

RESULTS

In order to demonstrate the usefulness of integrating GRNCOP2 in Cytoscape, the new app was used to tackle a novel use case for GRNCOP2: the analysis of crosstalk between pathways. In this regard, datasets associated with Alzheimer's disease (AD) were analyzed using GRNCOP2 app and other apps of the Cytoscape ecosystem by performing a topological analysis of the AD progression and its synchronization with the Ubiquitin Mediated Proteolysis pathway. Finally, the biological relevance of the findings achieved by this new app were evaluated by searching for evidence in the literature.

CONCLUSIONS

The proposed crosstalk analysis with the new GRNCOP2 app focused on assessing the phase of the Alzheimer's disease progression where the coordination with the Ubiquitin Mediated Proteolysis pathway increase, and identifying the genes that explain the signalling between these cellular processes. Both questions were explored by topological contrastive analysis of the GRNs generated for the GRNCOP2 app, where several facilities of Cytoscape were exploited. The topological patterns inferred by this new App have been consistent with biological evidence reported in the scientic literature, illustrating the effectiveness of using this new GRNCOP2 App in pathway analysis.

AVAILABILITY

The GRNCOP2 App is freely available at the official Cytoscape app store: http://apps.cytoscape.org/apps/grncop2.

Shared genes between Alzheimer's disease and ischemic stroke

Aims

Although converging evidence from experimental and epidemiological studies indicates Alzheimer's disease (AD) and ischemic stroke (IS) are related, the genetic basis underlying their links is less well characterized. Traditional SNP‐based genome‐wide association studies (GWAS) have failed to uncover shared susceptibility variants of AD and IS. Therefore, this study was designed to investigate whether pleiotropic genes existed between AD and IS to account for their phenotypic association, although this was not reported in previous studies.

Methods

Taking advantage of large‐scale GWAS summary statistics of AD (17,008 AD cases and 37,154 controls) and IS (10,307 IS cases and 19,326 controls), we performed gene‐based analysis implemented in VEGAS2 and Fisher's meta‐analysis of the set of overlapped genes of nominal significance in both diseases. Subsequently, gene expression analysis in AD‐ or IS‐associated expression datasets was conducted to explore the transcriptional alterations of pleiotropic genes identified.

Results

16 AD‐IS pleiotropic genes surpassed the cutoff for Bonferroni‐corrected significance. Notably, MS4A4A and TREM2, two established AD‐susceptibility genes showed remarkable alterations in the spleens and brains afflicted by IS, respectively. Among the prioritized genes identified by virtue of literature‐based knowledge, most are immune‐relevant genes (EPHA1, MS4A4A, UBE2L3 and TREM2), implicating crucial roles of the immune system in the pathogenesis of AD and IS.

Conclusions

The observation that AD and IS had shared disease‐associated genes offered mechanistic insights into their common pathogenesis, predominantly involving the immune system. More importantly, our findings have important implications for future research directions, which are encouraged to verify the involvement of these candidates in AD and IS and interpret the exact molecular mechanisms of action.

Whole-exome sequencing for variant discovery in blepharospasm

BACKGROUND

Blepharospasm (BSP) is a type of focal dystonia characterized by involuntary orbicularis oculi spasms that are usually bilateral, synchronous, and symmetrical. Despite strong evidence for genetic contributions to BSP, progress in the field has been constrained by small cohorts, incomplete penetrance, and late age of onset. Although several genetic etiologies for dystonia have been identified through whole-exome sequencing (WES), none of these are characteristically associated with BSP as a singular or predominant manifestation.

METHODS

We performed WES on 31 subjects from 21 independent pedigrees with BSP. The strongest candidate sequence variants derived from in silico analyses were confirmed with bidirectional Sanger sequencing and subjected to cosegregation analysis.

RESULTS

Cosegregating deleterious variants (GRCH37/hg19) in CACNA1A (NM_001127222.1: c.7261_7262delinsGT, p.Pro2421Val), REEP4 (NM_025232.3: c.109C>T, p.Arg37Trp), TOR2A (NM_130459.3: c.568C>T, p.Arg190Cys), and ATP2A3 (NM_005173.3: c.1966C>T, p.Arg656Cys) were identified in four independent multigenerational pedigrees. Deleterious variants in HS1BP3 (NM_022460.3: c.94C>A, p.Gly32Cys) and GNA14 (NM_004297.3: c.989_990del, p.Thr330ArgfsTer67) were identified in a father and son with segmental cranio-cervical dystonia first manifest as BSP. Deleterious variants in DNAH17, TRPV4, CAPN11, VPS13C, UNC13B, SPTBN4, MYOD1, and MRPL15 were found in two or more independent pedigrees. To our knowledge, none of these genes have previously been associated with isolated BSP, although other CACNA1A mutations have been associated with both positive and negative motor disorders including ataxia, episodic ataxia, hemiplegic migraine, and dystonia.

CONCLUSIONS

Our WES datasets provide a platform for future studies of BSP genetics which will demand careful consideration of incomplete penetrance, pleiotropy, population stratification, and oligogenic inheritance patterns.