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

Alzheimer’s disease (AD) is the most prevalent neurodegenerative disease among the elderly, which is characterized by memory impairment, language, visuospatial skills and other cognitive domains decline [1]. It is reported that women have higher overall incidence than men [2–9]. Sexes differ in genetic drivers [10], clinical severity [11], and neuropathological manifestations [12, 13]. Focusing on the sex-specific AD genetic drivers could transform the way treatments are developed and administered and lead to more personalized interventions.

The primary pathologies of AD is associated with neurofibrillary tangles (NFTs), amyloid-β (Aβ) plaque deposition, inflammation, synaptic alterations, and neurovascular amyloidosis [14]. The sexual dimorphism in brain structure, genetic background, inflammation, gliosis, and immune module are considered as important implications for mechanistic investigation of AD [15, 16]. Men often have a greater brain volume than women, and so are less sensitive to pathological changes [17], such as atrophy, ATP synthase, the mitochondrial proteome, a redox protein, and cytochrome oxidase [18, 19]. Additionally, preclinical studies have demonstrated a significantly relationship between estrogen and soluble Aβ levels in the brains of wild-type mice after ovariectomy [20]. Sex differences also present in immune modulation. Women usually have stronger neuro-inflammation and neuro-immune response than men [21]. Some sex-specific genes are associated with AD pathogenesis such as amyloid and tau. For instance, females have a stronger correlation between APOE and tau than males [22]. In the GWAS study of CSF AD biomarkers, the female-specific roles played by SERPINB1 in amyloidosis, OSTN, and CLDN16 in tau pathology have been observed [23]. Another research has confirmed that a male-specific ubiquitin-specific peptidase 9 is a positive regulator of MAPT, a protein associated with AD [24]. As a multifactorial disease, biomarkers are crucial for accurate diagnosis, as well as aiding in the understanding of disease mechanisms. Fluid biomarkers may differ based on sex, but only a few studies have focused on this. Further research is needed.

Recently, with the development of the high-throughput sequencing, bioinformatics analysis is widely applied to unveil underlying mechanisms such as biomarkers identification or molecular classification of diseases [25, 26]. Our study is to investigate the critical sex-associated differentially expressed genes (DEGs) and identify novel diagnostic biomarkers. We first downloaded microarray datasets for peripheral blood in AD and control samples from the GEO database. Stratifying the total sample into two groups distinguished by sex. We identified DEGs among males and females separately. Next, series of enrichment analyses, protein–protein interaction (PPI) analysis, an unsupervised hierarchical clustering analysis and support vector machine (SVM) were performed. We identified hub genes in males and females separately. Combined with sex and age, a 15-gene-based diagnosis model was constructed. This study provides more molecular insights into the sex differences in AD, and identifies candidate biomarkers for diagnosis (Fig. 1).

AD is a common disorder in the elderly and females are at higher risk of developing AD than males. Previous researches were usually based on CSF or brain tissue that may limit the application. Many studies have demonstrated that the blood can be used to detect disease-related changes [35–37]. Our study recruited blood tissue microarray data, and we integrated the differentially expressed genes in females and males separately.

Base on sex-specific DEGs, the KEGG pathways with smallest p-value enriched in females include oxidative phosphorylation, protein export, and collecting duct acid secretion. KEGG pathways enriched in males include ribosome, transcriptional misregulation in cancer, staphylococcus aureus infection, and NOD-like receptor signaling pathway. We discovered that abnormality of pathway is often associated with energy metabolism in females, whereas in males, the abnormality is primarily related to immune regulation. Females are more susceptible to estrogen's effects on energy homeostasis. Research has shown that estrogen in females plays a significant role in AD sex differences in preclinical and clinical studies [38, 39]. Female-specific pathways have also been found in a study of sex differences, but there were no male-specific pathways, possibly because tissue and ethnicities differ [40]. During AD pathogenesis, ribosome dysfunction can lead to altered translation, specifically in astrocytes [41, 42]. During immune challenge, male astrocytes produced more inflammatory molecules than female derived from cerebral cortex [43]. These studies show that males generally appear to have higher neuroimmune tone in the brain. However, the association between sex and these pathways involved in AD has not been studied. Additionally, analyzing the BP of male-specific DEGs showed that immune function, such as killing of cells of other organism, disruption of cells of other organism, antimicrobial humoral response, were stronger in AD samples.

To gain comprehensive insights, GSEA was conducted to identify potential pathways among males and females. As expected, male-specific pathway including toll like receptor signaling pathway and natural killer cell mediated cytotoxicity favours an immunological abnormalities. Next, we further detected the immune checkpoint molecules. Three of the eight immune checkpoint genes dysregulated in males, however, there were no differential gene expression in females. An immune checkpoint is a molecule that inhibits the immune system. Through controlling immune responses, they are essential to maintain self-tolerance and prevent autoimmune reactions. The observed sex dimorphism in immune checkpoint inhibitor responses is likely due to inherent differences between males and females [44–46]. Recent studies confirm that the immune checkpoint contributes to AD pathogenesis. It is well documented that mini-mental state examination (MMSE), tau proteins and Amyloid-β in AD were demonstrated to correlate with CTLA-4⁺T cells [47]. There could be sex-based differences in AD-intrinsic features that have been shaped by the immune system during AD development and immunoediting.

In the PPI network, SNRPG, RPS27A, COX7A2, ATP5PO, LSM3, COX7C, PFDN5, HINT1, PSMA6, RPS3A and RPL31 were identified as the female-specific hub genes, and SNRPG, RPL31, COX7C, RPS27A, RPL35A, RPS3A, RPS20 and PFDN5 were identified as the male-specific hub genes. SNRPG, RPS27A, COX7C, PFDN5, RPS3A and RPL31 were identified as common hubs. Nevertheless, other hub genes differ greatly between sexes. Moreover, the 13 genes were significantly dysregulated with AD progression.

A cluster of genes associated with the function and structural component of ribosomes, including RPS27A, RPL31, RPL35A, RPS3A, and RPS20, makes up most of the male-specific genes. Transcriptional changes in ribosome-related genes affect downstream translation in a broader way. Ribosome profiling was utilized in an in vitro experiment to investigate how the lack of oxygen and glucose immediately changes the transcription and translation of the genes in brain cells. The study concluded that oxidative stress has a greater effect on translation than transcription. It is alarming that testosterone can increase oxidative stress [48, 49] and oxidative stress-related conditions [50], and oxidative stress exacerbates testosterone's negative impact on cognition through its negative effects on androgens [51–54]. It can be inferred that sexually dimorphic differences may be a result of the expression of sex-specific ribosomal differential genes in AD. PSMA6 belongs to component of the 20S core proteasome complex that participates in the ATP-dependent degradation of ubiquitinated proteins. Ubiquitin proteasome system activity is integral to estradiol’s effects on memory, then this could lead to exciting new avenues of basic research into hormonal regulation of cognition, which could lead to important clinical implications for treating neurodegenerative disorders in which sex-based differences play a role [55]. LSM3 and SNRPG take role in the spliceosome's construction. As a component of the U4/U6-U5 tri-snRNP and the U1, U2, U4, and U5 snRNP complexes, they mostly performs pre-mRNA splicing-related tasks, including as RNA silencing and destruction. A snRNA(U1)accumulation has been observed in Alzheimer's disease [56]. In human brains, identifying cryptic splicing errors with neurofibrillary tangle burden implicates spliceosome disruption and transcriptome perturbation in AD Tau-mediated neurodegeneration [57]. These results show an independent function of U1 snRNA in regulating RNA splicing, suggesting that aberrant RNA processing may mediate neurodegeneration [58]. One study profiled the expression of snRNAs by applying small RNA sequencing to sncRNA isolated from anterior cingulate cortex of schizophrenia patients. Two snRNAs were found to be differentially expressed between female cases and controls [59]. However, no report has been published on the relationship between snRNA and sex differences in AD. HINT1 belongs to the histidine triad superfamily, which is widely expressed in different tissues. Ageing of the brain is a major risk factor for many neurodegenerative disorders including Alzheimer's. The downregulation of HINT1 has also been reported in diabetes and AD [60, 61]. Yu W et al. [62] found that potassium 2‐(1‐hydroxypentyl)‐benzoate can increase HINT1 expression levels, thus improving spatial learning and memory deficits in diabetic animals. Chen Q et al.'s study suggests that HINT1 may be associated with schizophrenia and the association is sex specific [63]. According to our study, HINT1 was found to be significantly differentially expressed in female AD. All these results suggest that HINT1 may play a role in neuronal function, but its exact physiological and cellular functions in AD remain unknown. PFDN5 belongs to prefoldin family. It is highly expressed in neurons and other neural cells, which can protect cells from apoptosis by decreases the toxicity of misfolded proteins. Interestingly, whole blood mRNA expression data from Alzheimer's patients revealed downregulation of PFDN5, which could be related to higher levels of toxicity of Aβ [64]. The results of our research analysis support this conclusion. The present study identified ATP5PO and COX7A2 as female-specific hub genes, which were significantly associated with ATP synthesis, heat production and oxidative phosphorylation. In female brains, estrogen control of glucose metabolism is dismantled during midlife, resulting in a shift in fuel systems and the emergence of dynamic neuroimmune phenotypes. As fuel use shifts, white matter is at risk of catabolism [65]. PSMA6 belongs to component of the 20S core proteasome complex that participates in the ATP-dependent degradation of ubiquitinated proteins. Ubiquitin proteasome system activity is integral to estradiol’s effects on memory, then this could lead to exciting new avenues of basic research into hormonal regulation of cognition, which could lead to important clinical implications for treating neurodegenerative disorders in which sex-based differences play a role [55]. From above analysis, it can be concluded that 13 genes were potential biomarkers for pathogenesis of AD. Assessing and confirming the clinical utility, we constructed a 15-marker panel (13 genes combined with age and sex) based SVM model, which can effectively predicts AD in independent testing sets.

Our work differs most significantly from previous work in the following ways: 1) Here, for the first time, we show the molecular mechanisms of sex differences in peripheral blood from Alzheimer's patients. 2) To identify the diagnostic value of these genes, we used machine learning methods to reduce overfitting and it can be improved in future studies.

Our study has some limitations: 1) Although we registered a second dataset for external verification, our results need to be confirmed by larger-scale studies. 2) The group differences in education and body mass index are potential confounds that could have affected our results, so further research is needed to examine the risks for these subgroups. 3) The mechanisms of these hub genes remain undefined in AD and they could be further explored to elucidate the functions and underlying mechanisms.

Our analyses revealed different hub genes of AD by the analysis of DEGs for males and females separately. We find that the pathophysiological pathways of AD differ in males and females. Using 13 genes as a base, we developed a diagnostic model with a high AUC value in peripheral blood This study provides insight into the underlying molecular mechanisms for sex dimorphism as well as potential biomarkers that may be useful for diagnostics and therapy.

Not applicable.

Wencan Ji designed the study and extracted the data. Wencan Ji, Ke An and Canjun Wang contributed statistical analyses. Wencan Ji and Shaohua Wang prepared figures and tables. Wencan Ji prepared the manuscript. Shaohua Wang approved the final draft. The author(s) read and approved the final manuscript.

This work was supported by the National Natural Science Foundation of China (Nos. 81570732 and 81870568, SW).

The datasets generated and/or analyzed during the current study are available in the [GSE63060↗] repository [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE63060↗], in [GSE63061↗] repository [https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE63061↗].