Bioinformatics analysis of diagnostic biomarkers for Alzheimer's disease in peripheral blood based on sex differences and support vector machine algorithm

Abstract

Background The prevalence of Alzheimer's disease (AD) varies based on gender. Due to the lack of early stage biomarkers, most of them are diagnosed at the terminal stage. This study aimed to explore sex-specific signaling pathways and identify diagnostic biomarkers of AD. Methods Microarray dataset for blood was obtained from the Gene Expression Omnibus (GEO) database of GSE63060 to conduct differentially expressed genes (DEGs) analysis by R software limma. Gene Ontology (GO) analysis, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis and Gene set enrichment analysis (GSEA) were conducted. Immune checkpoint gene expression was compared between females and males. Using CytoHubba, we identified hub genes in a protein–protein interaction network (PPI). Then, we evaluated their distinct effectiveness using unsupervised hierarchical clustering. Support vector machine (SVM) and ten-fold cross-validation were used to further verify these biomarkers. Lastly, we confirmed our findings by using another independent dataset. Results A total of 37 female-specific DEGs and 27 male-specific DEGs were identified from GSE63060 datasets. Analyses of enrichment showed that female-specific DEGs primarily focused on energy metabolism, while male-specific DEGs mostly involved in immune regulation. Three immune-checkpoint-relevant genes dysregulated in males. In females, however, these eight genes were not differentially expressed. SNRPG, RPS27A, COX7A2, ATP5PO, LSM3, COX7C, PFDN5, HINT1, PSMA6, RPS3A and RPL31 were regarded as hub genes for females, while SNRPG, RPL31, COX7C, RPS27A, RPL35A, RPS3A, RPS20 and PFDN5 were regarded as hub genes for males. Thirteen hub genes mentioned above was significantly lower in both AD and mild cognitive impairment (MCI). The diagnostic model of 15-marker panel (13 hub genes with sex and age) was developed. Both the training dataset and the independent validation dataset have area under the curve (AUC) with a high value (0.919, 95%CI 0.901–0.929 and 0.803, 95%CI 0.789–0.826). Based on GSEA for hub genes, they were associated with some aspects of AD pathogenesis. Conclusion DEGs in males and females contribute differently to AD pathogenesis. Algorithms combining blood-based biomarkers may improve AD diagnostic accuracy, but large validation studies are needed.

Research on roller bearing fault diagnosis method based on flexible dynamic adjustable strategy under data imbalance

“…With advancing computer and sensor technology, fault diagnosis methods based on data driven have been continuously proposed, and common machine learning methods include artificial neural network [4][5][6], random forest [7,8], extreme learning machine [9,10] and support vector machine (SVM) [11,12]. Among them, SVM, as a classic small sample learning algorithm, has been widely used due to its good sparsity and generalization ability in various fields such as image recognition, text classification, fault diagnosis and bioinformatics [13][14][15]. Unfortunately, the dual problem of SVM is a quadratic programming problem, which often takes up a lot of time in the process of solving quadratic programming problems and has low classification efficiency.…”