Added: Radley Opperman - Date: 15.11.2021 21:21 - Views: 25034 - Clicks: 7809
Methods: We took a meta-analysis approach to study gene expression in the brains of 1, AD Personal sex in Dobin am See and age-matched controls and whole blood from AD patients and age-matched controls in seven independent datasets. Sex-specific gene expression patterns were investigated through use of gene-based, pathway-based and network-based approaches.
Cell type deconvolution from whole blood gene expression data was performed to identify differentially regulated cells in males and females with AD. : Strikingly gene-expression, network-based analysis and cell type deconvolution approaches revealed a consistent immune ature in the brain and blood of female AD patients that was absent in males.
Interestingly, this gene expression program was missing in the brains of male AD patients. Interestingly, among males with AD, no ificant differences in immune cell proportions compared to controls were observed. Machine learning-based classification of AD using gene expression from whole blood in addition to clinical features produced an improvement in classification accuracy upon stratifying by sex, achieving an AUROC of 0.
AD prevalence increases dramatically with age, where the majority of cases are in individuals above the age of 65 Hebert et al. Although AD was identified more than a century ago Goedert and Spillantini,its cause and pathophysiology are not fully understood, and there are no available treatments that aid in halting or reversing the disease Cummings et al. Accordingly, it is of high priority to tackle AD, as it is projected to triple in incidence by as a consequence of population aging Riedel et al.
While the exact cause and pathophysiology remain unknown, a of mutations and genetic risk factors have been identified as associated with AD. ApoE is a lipid binding protein, that plays a central role in lipid transport and metabolism. It is highly expressed in the brain, and is important for maintaining neuronal membranes during inflammation and damage.
Sex is another major risk factor in AD. Female sex is associated with increased AD incidence, exacerbated pathophysiology and increased rate of cognitive decline related to the disease progression Andersen et al. It has been conjectured that the higher prevalence in females is a result of longer life span Carter et al.
Alternatively, studies have alluded to sex-specific hormonal and metabolic changes that interplay with the onset and progression of AD dementia Altmann et al. Despite the clear therapeutic potential to better understand these pathophysiological patterns, there is still little understanding of the mechanisms underlying sex-specific differences in AD.
With the rising prevalence of AD, it is critical to facilitate the development of robust means to detect AD early and discover therapeutic interventions Cummings et al. Technological innovations and the increasing availability of large transcriptomic datasets present worthwhile avenues to study and characterize the molecular underpinnings of AD stratified by sex. Here, we analyze publicly available gene expression datasets from over 1, brain and blood samples to characterize this highly complex disease.
To derive sex-specific transcriptomic molecular atures, we perform a meta-analysis, differential gene expression, weighted gene co-expression network analysis, pathway enrichment, and cell-type deconvolution in a large cohort of brain and blood samples from AD patients and healthy controls Figure 1.
We further characterize these atures and apply machine learning to build a predictive model based on biomarkers identified in the blood of AD patients. Our findings reveal sex-associated gene expression patterns in AD, which provide clinical implications for identifying more accurate, and less invasive biomarkers, as well as efficacious therapeutics tailored to better fit the complex molecular profiles in AD. Figure 1. Meta-analysis overview.
Diagram depicting the study overview including all datasets used and analyses performed. Datasets were merged using the ComBat package in R. WGCNA was used for network analyses. The linear SVM was trained to classify AD and control patients using the transcriptomic ature obtained via meta-analysis of blood studies. To minimize technical variability, brain samples were restricted to RNA-sequencing studies while blood analyses were restricted to microarray studies.
Brain samples were restricted to the hippocampus, parietal cortex, temporal cortex and prefrontal cortex. Meta-analysis was conducted separately for brain and blood studies according to standard quality control, normalization, and batch correction procedures.
All data processing was conducted using R v3. Alignment quality metrics were generated using Picard Counts-per-million CPM were calculated for all studies. Genes with missing gene length or GC content percentage metrics were removed. Library normalization was performed using conditional quantile normalization. Following read alignment and normalization, studies were merged using common genes between the four studies.
Mean value imputation was performed for missing gene expression values. Quantile normalization was performed across studies. The ComBat function from the sva package Leek et al. Principal component analysis PCA plots were generated to evaluate the success of batch correction and to detect outliers. Raw data were not available for the ADNI dataset and therefore normalized expression data were used for all studies.
Outlier removal was performed on individual studies by removing probes whose mean expression was outside 1. Probe IDs were mapped to gene symbols. Expression value of probes mapping to the same gene were reported as the median of all probes mapping to that gene Zhang, Principal component analysis PCA plots were generated to evaluate successful batch correction. Weiner, MD.
All differential gene expression analyses were performed separately for brain and blood samples. The Limma package Ritchie et al. An additional covariate of education was used in the blood analyses. Education was not available for all brain samples and therefore was not included as a covariate.
A cutoff false discovery rate FDR of 0.
Fold changes were calculated using the individual study data before merging and weighted by sample size. For blood analyses, a FC cutoff was not used to maximize gene discovery, given that we expect als to be considerably lower in the periphery than we do in disease tissue.
ificant overlap between up- and down-regulated genes between males and females was assessed using a hypergeometric test. Functional enrichment analysis of gene lists was carried out by overrepresentation analysis using the KEGG Kanehisa and Goto, database of biological pathways. All analyses were performed separately for brain and blood samples. In ed WGCNA, a module was defined as a set of genes whose expression is highly correlated in the same direction.
In brief, pairwise, ed similarity matrices were computed separately for male and female gene expression profiles. Pairwise similarity between two gene expression profiles, x i and x j was defined as:. The adjacency matrix Personal sex in Dobin am See transformed into a Topological Overlap Matrix as ly described Langfelder and Horvath, To identify clusters of interconnected genes, termed modules, hierarchical clustering was performed on Topological Overlap Matrix and modules were selected using the Dynamic Branch Cutting approach, as ly described Langfelder and Horvath, Module Z-summary scores were computed to assess module preservation between male and female networks, as described ly Langfelder et al.
A Z-summary score greater than ten was considered to be strong evidence of preservation between the two networks. A score between two and ten was considered to represent weak to moderate evidence of preservation, as ly described Langfelder et al. ificant modules were characterized by performing functional gene enrichment using the KEGG database of biological pathways Farrer, To identify central regulators of gene expression, we identified hub genes within ificant modules, as described ly Langfelder and Horvath, We also restricted hub genes to those that were differentially expressed in AD vs.
A gene expression profile of 22 reference cell populations was built using differential gene expression of purified or enriched cell populations from the authors of CIBERSORT. A cutoff FDR of 0. To assess the relative value of stratifying by sex in increasing model performance, we compared the performance of three models built using pooled male and female samples, male samples only, and female samples only.
For models with transcriptomic data, we included gene expression data from the corresponding sex. A random search over the space 10 —4 to 10 4 with five-fold cross validation was used to optimize the C hyper-parameter, or the degree of regularization penalty applied for misclassified points.
Receiver operating characteristic ROC curves were generated from the test set. Model performance was assessed using the area under the ROC curves. Feature importance was determined using the absolute value of the model coefficients corresponding to the vector coordinates orthogonal to the model hyperplane. Given the imbalance in the proportion of AD cases in the female samples compared to male samples, we down sampled our brain and blood datasets to assess whether our were primarily driven by differences in statistical power between males and females.
Specifically, we performed iterations of down sampling. In each iteration, we down sampled the female samples in our dataset such that the total of AD cases and controls was the same in the male and female groups. For example, the of cases and controls in the original dataset and down sampled dataset for the brain data are presented below:.Personal sex in Dobin am See
email: [email protected] - phone:(836) 887-8250 x 3689
Dobin Am See Buddhist Dating