Identification of biomarkers and pathways for the SARS-CoV-2 infections in obstructive sleep apnea patients based on machine learning and proteomic analysis

During the past three years, an extensive number of individuals have faced the challenging period of coronavirus disease outbreaks. The latest estimates from the World Health Organization indicate a staggering 765,222,932 confirmed cases of COVID-19 as of May 3rd, 2023, with 6,921,614 reported fatalities (https://​covid19.​who.​int/​). This has placed a significant burden on healthcare systems worldwide. Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the causative agent of COVID-19, primarily affects the upper respiratory tract and lung tissues. Previous studies have identified older age, diabetes, severe asthma, hypertension, obesity, and cardiovascular and cerebrovascular comorbidities as high-risk factors associated with COVID-19 infection and unfavorable prognosis [1‐3]. Intriguingly, several of these comorbidities, such as hypertension, diabetes mellitus, obesity, and vascular disease, have been shown a strong association with OSA [4‐6]. Hence, exploring shared pathogenic mechanisms between these two diseases has captured significant interest among researchers and clinicians.

OSA is a disorder characterized by recurring episodes of pharyngeal airway collapse, leading to intermittent partial or complete cessation of breathing during sleep. This condition has been strongly associated with various cardiovascular, cerebrovascular, and metabolic disorders. In recent years, extensive clinical observational studies have indicated that individuals with OSA may be more susceptible to COVID-19 infection and experience poorer outcomes [7‐15]. Furthermore, two meta-analyses have revealed that OSA is associated with a 2.0 and 1.7 times increased risk of severe COVID-19, respectively [16, 17].

Nevertheless, the precise pathological mechanisms underlying the association between OSA and the susceptibility to COVID-19 infection as well as the development of COVID-19 complications remain inadequately elucidated. Several investigations propose an intricate interplay involving intermittent hypoxia, oxidative stress, sympathetic activation, inflammation, endothelial dysfunction, and associated complications [17, 18]. To delve further into these intricate mechanisms, a case–control study has unveiled a correlation between sleep-related hypoxemia and an augmented likelihood of adverse COVID-19 outcomes. This discovery implies a potential linkage between COVID-19-induced hypoxia and pre-existing sleep-related hypoxemia. Furthermore, given the concurrent involvement and impact on the respiratory system in both conditions, certain studies postulate that individuals with pre-existing OSA are at an elevated risk of morbidity and mortality upon exposure to SARS-CoV-2 due to an excessive inflammatory response [19, 20]. Moreover, a review conducted by Mashaqi et al. suggested that alterations in the gut microbiota, which occur in the context of COVID-19 and OSA, can exacerbate systemic inflammation by facilitating the translocation of microorganisms across a compromised intestinal barrier [8]. Overall, the current theories and hypotheses fail to provide a comprehensive elucidation of the intrinsic molecular mechanisms underlying the association between OSA and COVID-19. Comprehending the potential interconnections and plausible molecular pathways connecting these two conditions is imperative for the effective management of COVID-19 patients with concomitant OSA and for the identification of viable therapeutic interventions.

In recent times, the clinical utilization of bioinformatics has witnessed a surge, enabling the exploration of intricate disease mechanisms. Its remarkable advantages lie in its capacity to prognosticate the structure, prognosis, functionality, and evolution of uncharted genes and proteins, while also facilitating the screening of medications and vaccines from extensive repositories of sequence data. Consequently, the objectives of this investigation encompassed an examination of the plausible association between COVID-19 and OSA, along with an in-depth exploration of the molecular mechanisms that underpin this comorbidity, employing a bioinformatics framework.

Various observational studies have indicated that individuals afflicted with OSA exhibit a heightened vulnerability to severe SARS-CoV-2 infection [10, 15‐17], albeit this association diminishes upon adjustment for other comorbidities (such as obesity and cardiovascular disorders) [33]. A Mendelian randomization study also yielded inconclusive evidence regarding the causal relationship between OSA and COVID-19. Conversely, a population-based study demonstrated that even after accounting for the effects of other comorbidities (e.g., obesity), OSA remains capable of doubling the risk of COVID-19-related adverse events, including hospitalization and death. Consequently, further comprehensive investigations are imperative to elucidate the intrinsic relationship between OSA and COVID-19. Within this study, fourteen feature genes shared by COVID-19 and OSA have been identified. Subsequently, through the construction of a protein–protein interaction network incorporating these common feature genes, we have identified 10 central genes (TP53, CCND1, MDM2, RB1, HIF1A, EP300, STAT3, CDK2, HSP90AA1, and PPARG). The subsequent analysis proceeded with functional enrichment analysis, TF-gene interactions, TF-miRNA interaction networks, and drug candidate screening.

Protein–protein interaction is the process by which two or more protein molecules form a protein complex through non-covalent bonds. At the molecular level, viruses invade cells and use a complex network of protein–protein interactions to manipulate cellular processes, and host cells initiate activation of innate antiviral defenses and adaptive immune systems to control viral replication through complex protein–protein interactions. Therefore, knowledge of protein–protein interactions is essential for understanding the virus-host relationship. Zhang et al. identified relevant genes in the virus-human protein interaction network, including UBL4A, GNB1, UMPS, POTEF, UBL4B, UBBP4, PARK2, PTPN11, and SCOC based on a random walk model [34]. In particular, Yang et al. developed the Human-Virus Interaction DataBase (HVIDB, http://​zzdlab.​com/​hvidb/​), which can accurately predict interactions between human host and viral proteins [35].

Currently, integrated machine learning based on multiple feature selection algorithms has been widely used to find biomarkers for diseases, which can screen potential biomarkers with high sensitivity, accuracy, and stability, and construct efficient and stable diagnostic models. Since the outbreak of COVID-19, numerous studies have analyzed transcriptomic, proteomic, metabolomic, and epigenomic data from COVID-19 patients based on machine learning to find reliable and stable COVID-19 biomarkers. Guo et al. developed a random forest machine learning model based on proteomic and metabolomic data from COVID-19 patients, which incorporated genes such as SAA2,ALB,CRP,SAA1,HABP2, and HP, and was able to discriminate well between non-severe and severe COVID-19 patients (AUC = 0.957) [36]. Using LASSO and ridge regression penalties, Zhou et al. identified a biomarker combination containing ORM1/AGP1, ORM2, FETUB, and CETP that demonstrated excellent reliability in distinguishing COVID-19 patients from healthy volunteers [31]. Cai et al. identified five biomarkers of COVID-19 by applying Boruta and minimum redundancy maximum relevance methods, which were PSMB8, COLCA2, FAM83A, LGALS3BP and IRF9. In addition, Cai et al. analyzed the methylation dataset of COVID-19 by utilizing the Monte Carlo feature selection method analyzed the methylation dataset of COVID-19 and found that EPSTI1, NACAP1, SHROOM3, C19ORF35 and MX1 as key features could significantly distinguish COVID-19 patients from non-COVID-19 patients [37, 38].

Prior investigations have identified five of these central genes (TP53, HIF1A, STAT3, HSP90AA1, PPARG) to be linked to the pathogenesis of both COVID-19 and OSA. The tumor suppressor gene TP53, a crucial participant in the apoptotic signaling cascade, assumes a significant role in monitoring cellular division and possesses reparative effects on aberrant division and cellular damage. Multiple studies have revealed a downregulation of TP53 gene expression in patients infected with SARS-CoV-2, consequently fostering viral replication and increasing the likelihood of tumorigenesis [39‐42]. Intriguingly, periodic intermittent hypoxia has been implicated as a causative factor for non-small cell lung cancer in TP53fl/fl mice, thereby facilitating accelerated tumor growth [43]. Hence, we postulate that individuals afflicted with the concomitant presence of these two ailments are predisposed to a heightened risk of cancer or tumor development, ultimately resulting in diminished overall survival rates.

HIF1A (hypoxia-inducible factor 1-alpha), serves as a prominent regulator of cellular and systemic homeostatic responses to hypoxia by orchestrating the transcriptional activation of numerous genes. These genes encompass those involved in energy metabolism, angiogenesis, apoptosis, and others whose protein products enhance oxygen delivery or facilitate metabolic adaptation to hypoxic conditions [44]. Notably, previous investigations have indicated that the prognosis of COVID-19 is significantly influenced by the cytokine storm triggered by SARS-CoV-2. Transcriptome analyses have unveiled that SARS-CoV-2 infection induces mitochondrial damage and the generation of mitochondrial-derived reactive oxygen species (Mito-ROS), thereby promoting the expression of HIF-1A and consequent SARS-CoV-2 infection and cytokine production [45, 46]. Furthermore, HIF-1A assumes a vital role in regulating metabolic pathways and inflammatory processes [47]. The dysregulation of the HIF-1A pathway contributes to the development of various diseases, including cancer, cardiovascular disease, and Alzheimer's disease [48‐51]. Importantly, many of these disorders have been demonstrated to be associated with severe SARS-CoV-2 infection, thereby suggesting the prognostic impact of HIF-1A on COVID-19 [52]. On the other hand, numerous studies have illustrated the close association between HIF-1A and OSA, given that OSA is characterized by chronic intermittent hypoxia (CIH). CIH stimulates the upregulation of the hypoxia-inducible transcription factor HIF1A, contributing to the independent risk of OSA for atherosclerosis and heightened cardiovascular mortality. Mechanistically, hypertension, dyslipidemia, insulin resistance, systemic inflammation, and oxidative stress are among the underlying mechanisms through which CIH accelerates atherosclerotic disease [53]. A recent study has demonstrated that mice exposed to CIH exhibited impaired clearance of triglyceride-rich lipoproteins. Subsequent investigations have suggested that this phenomenon may be attributed to the transcriptional regulation of HIF-1A, which upregulates the expression of the lipoprotein lipase (LPL) inhibitor (Angptl4), thereby inactivating LPL and impeding lipid clearance [54]. Additionally, CIH exacerbates non-alcoholic fatty liver disease (NAFLD) by inducing hepatic oxidative stress via the activation of the HIF-1A gene [55]. Furthermore, HIF-1 has been implicated in the pathological processes of hypoxia and has been associated with various other diseases such as diabetes mellitus, idiopathic pulmonary fibrosis, pulmonary arterial hypertension, systemic hypertension, myocardial injury, and cognitive deficits [56‐61]. In summary, the myriad metabolic and pathophysiological responses elicited by intermittent hypoxia in individuals with OSA may contribute to their susceptibility to COVID-19 and portend a poorer prognosis.

Inflammation constitutes a hallmark feature of both OSA and SARS-CoV-2 infections. Hypotheses suggest that alterations in the gut microbiota, which are associated with COVID-19 and OSA, may compromise the integrity of the intestinal barrier, thereby facilitating the translocation of bacteria into the systemic circulation and subsequently promoting systemic inflammation [8]. Notably, systemic inflammation is particularly pronounced in the context of COVID-19. STATs (signal transducer and activator of transcription) are well-known components of the JAK/STAT signaling pathway. Previous investigations have revealed that the viral component of SARS-CoV-2 induces malfunctioning of STAT1 and compensatory overactivation of STAT3. Consequently, a positive feedback loop between STAT3 and plasminogen activator inhibitor-1 (PAI-1) is established in SARS-CoV-2-infected cells, thereby perpetuating a cycle of activation that triggers the release of pro-inflammatory cytokines and chemokines [62].

In a transcriptomic analysis, the connection between HSP90A family class A member 1 (HSP90AA1) and the quantity of SARS-CoV-2 RNA was unveiled [63]. Moreover, HSP90AA1 was identified as a risk factor for the coexistence of COVID-19 and cardiovascular disease [64]. Notably, HSP90 levels were significantly elevated in patients with OSAS when compared to control subjects [65]. These findings suggest that HSP90 may potentially play a role in the progression of both of these diseases.

PPARG (peroxisome proliferator-activated receptor gamma), a crucial transcription factor, governs both inflammation and fatty acid metabolism. Studies have demonstrated that PPARG agonists can enhance the expression of ACE2, thereby contributing to the invasion of SARS-CoV-2 [66]. Furthermore, OSA is characterized by metabolic dysfunction and obesity, and the PPARG gene is implicated in the regulation of lipid metabolism and glucose homeostasis. This suggests that PPARG may also play a potential role in the pathophysiological mechanisms underlying the comorbidity of COVID-19 and OSA [67]. As for the other five central genes (CCND1, MDM2, RB1, EP300, CDK2) in the study, their associations with OSA patients have not been reported. Therefore, future investigations could delve into exploring the potential mechanisms that underlie the comorbidity between COVID-19 and OSA from additional perspectives.

Condensed chromosomes, centromeric regions, autolysosomes, and molecular functions related to DNA insertion or deletion was enriched in this study. The primary constriction along a condensed chromosome is known as the centromere region, and the end of the chromosome as the telomere. These regions are essential for the proper integrity and segregation of chromosomes, as well as for the hierarchy of order within the cell nucleus. A study conducted at the Federal University of São Paulo in Brazil shows that the telomere shortening that naturally occurs with aging and is accelerated by OSA can be mitigated by the use of continuous positive airway pressure, indicating that accelerated telomere shortening due to OSA can therefore lead to premature cell aging [68]. It is now well established that SARS-CoV-2 can enter host cells via the ACE2 receptor. Interestingly, the gene encoding ACE2 is located on an X chromosome, and according to Lyon's theory, one of the two X chromosomes is transcriptionally silenced, a process that results in the condensation of an X chromosome into a compact structure known as a barr body [69, 70]. Autolysosome is formed when an autophagosome or phagosome fuses with a lysosome [71]. As a marker of autolysosomal degradation, P62 protein expression in blood cytotoxic T cells was increased in OSA patients compared to primary snoring patients, indicating that OSA patients had impaired autophagy activity [72]. In COVID-19, the SARS-CoV-2 protein ORF3a affects autolysosome formation by hindering the assembly of the SNARE complex, and furthermore, SARS-CoV-2 infection-induced autolysosome accumulation facilitates progeny virus propagation [73‐75]. As for DNA insertion or deletion, it was shown that angiotensin-converting enzyme gene insertion/deletion polymorphisms differed significantly in OSA disease groups with different severities and may be involved in the pathogenesis of hypertension [76, 77]. In COVID-19, insertion-and-deletion mutations in several genes including Nsp2, Nsp3, S1, and ORF8 genes have been observed, which may provide further insights into the developmental process of SARS-CoV-2 [78, 79].

The study revealed that gas homeostasis may be a crucial shared pathogenic mechanism in COVID-19 and OSA, as indicated by the analysis of GO enrichment. Gas homeostasis refers to the tightly regulated maintenance of oxygen and carbon dioxide levels in the human body, involving multiple pathways. Recent research has suggested the involvement of these pathways in the pathogenesis of both COVID-19 and OSA. Hypoxia-inducible factors (HIFs) are known regulators of cellular responses to low oxygen levels. We have discussed in detail the significant role of HIF1A in the development of COVID-19 and OSA patients. Furthermore, OSA patients experience chronic intermittent hypoxia, while the increased production of pro-inflammatory cytokines in COVID-19 also affects gas exchange and disrupts gas homeostasis [80]. Oxidoreductase activity refers to the enzymatic activity involved in redox reactions, which is closely associated with oxidative stress [81]. It is increasingly recognized that COVID-19 pathogenesis and immunopathogenesis are closely linked to oxidative stress and cytokine storm. Oxidative stress contributes to the development of endothelial dysfunction, while the cytokine storm exacerbates this effect. Ultimately, these events trigger the activation of the blood clotting cascade, leading to blood coagulation and microvascular thrombosis. In the case of OSA, intermittent hypoxia and reoxygenation during sleep-disordered breathing also induce oxidative stress and inflammation[82, 83]. This resulting oxidative damage further contributes to the development and worsening of complications associated with OSA, including obesity, metabolic dysfunction, cardiovascular diseases, and cognitive impairment [82, 84, 85]. DNA repair and replication pathways play vital roles in maintaining the stability of the genome and ensuring cell survival. Recent studies have revealed that these pathways may also be involved in the pathogenesis of COVID-19 and OSA. Emerging evidence suggests that viral infection can induce DNA damage and trigger an altered DNA damage response. Specifically, certain products of SARS-CoV-2, such as ORF6 and NSP13 proteins, disrupt the DNA damage response kinase CHK1 and reduce the production of deoxynucleoside triphosphate (dNTP), which consequently leads to DNA damage response. In the case of OSA, increased levels of reactive oxygen species and repeated hypoxic conditions also disrupt DNA repair mechanisms and contribute to enhanced DNA damage [86]. In summary, gas homeostasis, oxidoreductase activity, and DNA repair and replication pathways may contribute to the shared pathogenesis of COVID-19 and OSA. Further studies are required to fully understand the precise roles of these pathways in these diseases and to explore their potential as therapeutic targets.

In this study, ten drug candidates were identified. To further strengthen the credibility of the results, we searched databases of information about drugs for the treatment of COVID-19, including DockCoV2 and DrugBank, assembled by Huang et al. [87‐90]. Of these, Topotecan and Estradiol had docking scores of -16.8 and -10.7, respectively, indicating potential as therapeutic agents.

It is important to acknowledge the limitations of this study. Firstly, the analysis was conducted using data from patients diagnosed with COVID-19 and OSA, rather than individuals specifically diagnosed with both disorders. This may impact the generalizability of the findings to the co-occurrence of COVID-19 and OSA. Additionally, it should be noted that the identified genes were not experimentally validated in OSA samples.

In conclusion, this investigation revealed a set of genes, molecular networks, and signaling pathways that are associated with both COVID-19 and OSA. These findings provide insights and potential avenues for further research into the molecular mechanisms underlying these two diseases. The study highlights a shared regulatory scheme and signaling mechanism between COVID-19 and OSA.