Previously published literature indicates that miRNAs are an important regulatory mechanism for AD-associated gene expression [14]. So far, several miRNAs have been shown to regulate AD-associated genes [14]. As the main miRNA mechanism of action is the downregulation of target genes, it is important to assess whether they are being over- or under-expressed in patients. Additionally, miRNA expression can be tissue-specific or bound to a specific mechanism, such as the regulation of extracellular vesicles, which are involved in cell communication [116]. As the understanding of the AD genetic background is not yet complete, observing miRNAs as a contributing factor may prove valuable.

The study results identified MTI subnetworks of varying sizes. The largest MTI subnetwork identified consists of 18 miRNAs and 15 target genes. The most prominent miRNA targets in the MTI network are APP, BACE1, NCSTN, SIRT, and SP1 as they are the targets of multiple miRNAs. The six smaller networks are composed of three to five nodes with two to four MTIs. In one subnetwork, two miRNAs regulate the same gene; where hsa-miR-302a-3p and hsa-miR-200c-3p both target PTEN. From Figure 1, it is also apparent that there are 12 MTI pairs not connected to the other subnetworks.

The methods used for the validation of miRNA–target interactions are not all equally reliable and have different validation statuses. The miRTarBase methods are divided into strong and less strong based on the validation status. Methods such as Western blot, qPCR, and reporter assay are considered to give more reliable information and are marked as methods with strong validation status. Microarrays, NGS, pSILAC, and other methods are, by contrast, considered to generate less strong evidence. Consensus on the validation strength of methods has not yet been achieved as studies use different definitions of what constitutes strong and less strong evidence when it comes to MTIs. The edges between targets and miRNAs in Figure 1 do not distinguish between strong and less strong validation. In the future, these data could be accounted for in the graphical network.

As previously mentioned, five genes had the largest number of edges in the MTI network. Among these is the BACE1 gene, regulated by seven AD-associated miRNAs in the MTI network. Beta-secretase 1 is a protease encoded by the BACE1 gene. Its main role is the extracellular cleavage of the amyloid precursor protein (APP). It cleaves APP into two components, one of which is known as C99. This component is then further cleaved by γ-secretase, releasing an amyloid beta peptide (Aβ), which is the primary component in amyloid plaques. These plaques are commonly found in the brains of AD patients. Due to the correlation of amyloid plaque formation and AD, BACE1 has been closely studied. Inhibiting BACE1 would prevent the formation of Aβ, and it has been speculated that BACE1 inhibitors may prevent the development of the disease [117].

Another miRNA target with multiple MTIs is APP. Like BACE1, APP is also regulated by seven miRNAs in the MTI network. As seen in Figure 1, BACE1 and APP are both targeted by miR-16. As previously mentioned, APP is cleaved by proteases and is the precursor for Aβ. Despite being of great interest in connection to AD, its function is not completely known. Evidence shows that increased APP expression could promote the production of Aβ, leading to a negative impact on neurons and synapses [118].

Nicastrin is a protein encoded by the NCSTN gene. NCSTN is targeted by four miRNAs in the MTI network, making it the gene with the third-highest number of interactions with AD-associated miRNA. It is part of the γ-secretase protein complex and is thus connected to the formation of Aβ. No solid evidence has so far been found that would connect it to the development of AD, though this does not exclude it as a potential contributing factor [119]. As seen in Figure 1, miR-16 regulates three highly interconnected genes in the MTI network—APP, BACE1, and NCSTN.

The SIRT1 gene encodes the enzyme NAD-dependent protein deacetylase sirtuin-1 (SIRT1). In the MTI network, SIRT1 is a target for two miRNAs: miR-181c and miR-9.

The roles and functions of human sirtuins are still largely unknown; however, SIRT1 has a known interaction with hypoxia-inducible factors 1α and 2α (HIF1A and EPAS1, respectively). HIF-1α and EPAS1 are important for proper brain development as they are crucial for cell adaptation to hypoxia [120]. In murine models, all gene expression alterations in EPAS1-deficient mice have previously been associated with AD and memory loss [121]. SIRT1 was found to deacetylate the tau protein in some cell cultures. Among other interactions, it was also observed that it has a protective role in microglia-dependent Aβ toxicity [122].

SP1 is a transcription factor and is targeted by two miRNAs in the MTI network: miR-29b and miR-375. SP1 may be involved in the development of AD as it can regulate the expression of several genes previously associated with AD, such as APP and tau protein genes. SP1 has also been shown to be significantly upregulated in the frontal cortex of AD patients [123].

PTEN is a gene that translates into the phosphatase and tensin homolog (PTEN) protein. PTEN is a target of miR-200c and miR-302a in the MTI network. Mutations in this gene are primarily associated with different types of cancer; however, they are also associated with AD through its role in synaptic and cognitive functions [124]. The gene also acts as a tumor suppressor, and its involvement with AD has been studied in mice [125].

Along with an MTI network, we also conducted a PPI analysis for AD-associated miRNA targets (Figure 2). Both the PPI and MTI network contained nodes with a large number of interactions, either with other targets (proteins) or miRNAs. The two networks share some nodes with multiple interactions. APP, BACE1, PTEN, SIRT1, and SP1 are targets that have the highest number of interactions in both networks. However, the PPI networks also include other highly connected proteins, such as STAT3, CDKN2A, E2F1, MFN2, RB1, and IGF1. The PPI network is highly interconnected: 41 of the 43 genes currently have at least one known or predicted interaction within the network. This level of interconnectedness of miRNA targets further points to the complex nature of AD.

A total of 37 AD-associated miRNAs were identified to be enriched in 68 biological pathways using the mirPath tool. For these 68 pathways, we conducted a secondary review of previously published literature and identified 44 pathways which have previously been described in association with AD. These 44 AD-associated pathways also include five pathways associated with other diseases: glioma, acute myeloid leukemia, colorectal cancer, hepatitis B, and type II diabetes mellitus. For example, glioma and AD share some common biological pathways associated with their development as well as genetic and environmental risk factors but are as of yet not causally related [75]. Common pathways with AD were also identified for acute myeloid leukemia [126] and colorectal cancer [80]. Interestingly, research has discovered an inverse correlation between AD risk and lung cancers [127] as well as other types of cancers [128,129,130]. Further studying the shared pathways between these diseases may yield additional insight into the role played by individual pathways in the development of AD.

Among the 44 AD-associated pathways, 39 pathways were not associated with diseases. These pathways are involved in inter- and intracellular signaling, gene regulation, cell adhesion, endocytosis, phagocytosis, and inflammation, which is expected, as these mechanisms have been shown to be involved in AD pathology [131]. The results of the analysis indicate that AD-associated miRNAs are involved in a variety of biological pathways. Based on the number and variety of pathways miRNA target genes are enriched in, miRNA appear to play a role in AD on multiple levels. Observing the disease at its endpoint, however, as is the tendency of study designs for AD-associated factors, has drawbacks. Due to the complexity of cellular regulatory mechanisms, observing dysregulations at the end point of a disease may not necessarily answer questions regarding its etiology. The interconnected nature of biological pathways means that the dysregulation of one pathway can cause the dysregulation of a second pathway. Though the second pathway is now dysregulated, studies observing the end-point of a disease’s progression will not be able to discern between the cause and effect [132]. Therefore, longitudinal studies spanning from the preclinical stage of disease development to its endpoint are vital for the understanding of AD and consequently identifying therapeutic approaches.

MTIs acquired from miRTarBase do not include all genes whose variants are commonly associated with increased risk of AD. Their absence may be due to the incomplete initial data set, the current lack of knowledge about the role of miRNA in the contribution to AD development, or an indirect mechanism through which these risk variants contribute to the disease. Some genes commonly associated with AD, such as APOE, are currently not included in the database. APOE is considered one of the most influential genetic risk factors for late-onset AD—specifically, one of its three major isoforms, APOE-E4. The full extent of interactions between APOE and miRNA is not yet understood; however, it has been shown that levels of miR-1908 were negatively correlated with APOE expression [133]. Other known genetic AD risk factors include PSEN1 and PSEN2, specifically for early-onset AD [134]. PSEN1 is regulated by miR-193a [135] while PSEN2 knockout microglia cells exhibited the downregulation of miR-146 [43]. Further research into the involvement of APOE, PSEN1, and PSEN2 is required in order to acquire a better understanding of their role in AD. Future research is needed to reveal complete understanding of the role of miRNA in APOE, PSEN1, and PSEN2 regulation.

Different methods have been used both for the identification of novel biomarkers and for diagnostic purposes. For example, a cell culture reporter assay was used to determine that miR-107 regulates the expression of BACE1 [57]. Cell culture reporter assays, ELISA, xMap Luminex, shotgun proteomics, and other methods are commonly used to perform biomarker assays. The methodology on AD biomarker detection, however, is not entirely consistent among laboratories. For example, individual laboratories have different concentrations of Aβ that are considered low or high for the purpose of assays [136]. Blennow and Zetterberg (2018) have evaluated a large number of studies on miRNAs associated with AD and highlighted the need for a standardized analytical protocol among research centers [2]. A standardized approach for determining whether AD-associated molecules are present in low or high concentrations for the purpose of diagnostics is necessary for reliable, reproducible studies on AD.

Despite several important contributions to the development of the study field, there are also some limitations inherent to the present study. The study is limited by data available in the miRTarBase and mirPath databases, as these databases do not include all miRNAs and their targets currently known to be associated with AD. MirTarBase is one of the most extensive miRNA databases, but due to the rapid pace of the developments in this field, it may be challenging to keep a database up to date. Additionally, our initial dataset did not include expression levels and tissue specificity for miRNA and targets. The stage of AD during which measurements were taken was also not included, though these data could be taken into account in the future. The focus of the present study is the identification of interactions between known AD-associated miRNA and targets, their visualization, and the analysis of their enrichment in biological pathways. Through this, the study contribution is an overview of the interplay between miRNAs and AD-associated genes.

miRNAs have been extensively studied for their use as AD biomarkers in previously published literature [137]. A study by Lugli et al. (2015) assessing exosomal miRNAs as potential AD biomarkers has observed the differential presence of miRNAs in the plasma of AD patients. Twenty miRNAs showed notable differences, and seven of those were used for AD status prediction of patients using machine learning. The machine learning model predicted the patient’s AD status based on samples with an 83–89% accuracy; however, the authors recommend a replication with a larger cohort. The addition of miRNA expression data into other AD biomarker diagnostic tests is likely to further increase the diagnostic accuracy [138]. The results of a study by Leidinger et al. (2013) showed that a panel of 12 blood-based miRNAs can be used to differentiate between AD patients and healthy controls with 93% accuracy. This panel of 12 miRNAs can also differentiate between AD and other CNS disorders with 74–78% accuracy [139]. Other studies testing circulatory miRNAs as biomarkers have also shown 75–95% accuracies in identifying AD [140]. These studies are, however, focused on the late stages of AD. Integrating blood or plasma miRNA biomarkers with other biomarkers, such as Aβ40 and Aβ42, are likely prospective methods of early disease detection.

In the present study, miR-9, miR-16, miR-34a, miR-106a, miR-107, miR-125b, miR-146, and miR-181c were miRNAs present in the highest number of MTIs in the network. These miRNAs present promising components of regulation of AD-associated genes. miR-16 and miR-34a are involved in processes key to the amyloid cascade hypothesis model of AD development [14]. miR-16 inhibits APP expression while miR-34a inhibits the expression of proteins connected with Aβ clearance [14]. miR-125b, meanwhile, is involved in the tau cascade hypothesis model, where its role is the inhibition of kinases responsible for tau hyperphosphorylation [14]. These miRNAs are differentially expressed in the brain and CSF; however, as diagnostic methods aim towards blood-based biomarkers, studies are necessary to elucidate whether they are also viable as circulatory biomarkers. Studies have shown the potential for blood-based miRNA biomarkers [137,138,139,140]. However, studies involving miRNA as circulatory biomarkers have, in the majority of cases, been performed with participants at the dementia stage of AD. Longitudinal studies with larger sample sizes are necessary to identify a combination of robust early detection biomarkers. As for therapeutic targets, research in the topic remains incomplete, being performed predominantly on cell cultures and murine models. Anti-miRNA (AM) approaches are challenged by an imperfect drug delivery system and unwanted effects on the expression of non-AD-associated genes, due to the multi-target nature of miRNAs. An anti-miR-146a (AM-146a) approach has shown a regained miRNA-associated homeostasis in murine models of AD. In cell cultures, AM-34a has returned overexpressed TREM2 [141] and SHANK3 [21] levels back to expected, normal levels, and thus, homeostasis. miRNAs as therapeutic targets should therefore not be ruled out, but more research is necessary to identify their level of potential for this purpose.

miRNAs play a prominent role in the regulation of AD-associated gene expression, with vast research potential into targets for screening, diagnosis, or treatment. Our analysis revealed seven subnetworks of MTIs, representing disease modules, which have the potential for network-based biomarker development. Further investigation into the cause of the upregulation or downregulation of miRNA may also prove useful in the search for the cause of AD. As there are a large number of miRNAs to consider in AD development, research, or screening, lab-on-a-chip technology is likely to be an efficient and cost-effective method to utilize.