Non-coding RNAs in Alzheimer’s disease: perspectives from omics studies

Abstract

Neurodegenerative diseases such as Alzheimer’s disease (AD) are characterized by the progressive loss of neurons in the brain and the spinal cord. The pathophysiology of AD is multifactorial with heterogeneous molecular manifestations. The lack of efficacious therapies for AD reinforces the importance of exploring in depth multifaceted disease mechanisms. Recent progresses on AD have generated a large amount of RNA-sequencing data at both bulk and single cell levels and revealed thousands of genes with expression changes in AD. However, the upstream regulators of such gene expression changes are largely unknown. Non-coding RNAs (ncRNAs) represent the majority of the human transcriptome, and regulatory ncRNAs have been found to play an important role in regulating gene expression. A single miRNA usually targets a number of mRNAs and thus such ncRNAs are particular important for understanding disease mechanisms and developing novel therapeutics. This review aims to summarize the recent findings on the roles of ncRNAs in AD from ncRNA-omics studies with a focus on ncRNA signatures, interactions between ncRNAs and mRNAs, and ncRNA-regulated pathways in AD. We also review the potential of specific ncRNAs to serve as biomarkers and therapeutic targets for AD. In the end, we point out future directions for studying ncRNAs in AD.

Introduction

Neurodegenerative diseases are characterized by the progressive loss of neurons in the brain and the spinal cord (1). In 2021, over 10 million Americans older than 65 years were living with a neurodegenerative disability, whereas two-thirds of these patients present Alzheimer’s disease (AD)-like symptoms. The pathophysiology of AD is multifactorial with heterogeneous molecular manifestations.

Although some important hallmarks such as amyloid-beta (Aβ) pathology and neurofibrillary tangles have been extensively characterized in AD, increasing evidence suggests that the dysregulation of protein and lipid metabolism, neuroinflammation and immune processes, endo-lysosomal pathway, oxidative stress, synaptic processes and myelin formation may contribute to the pathogenesis of AD (2–6). The lack of efficacious therapies for AD reinforces the importance of exploring in depth multifaceted disease mechanisms.

Non-coding RNAs (ncRNAs) represent the majority of the human transcriptome (7); for a long time, ncRNAs were considered junk transcripts (8). However, in recent years, the advances in cutting-edge molecular profiling technologies allow us to determine the biological relevance of ncRNAs (7). ncRNAs are divided into two main categories according to their functions: (i) housekeeping ncRNAs and (ii) regulatory ncRNAs. The housekeeping ncRNAs are responsible for mediating cellular homeostasis; for example, ribosomal RNAs and transfer RNAs are essential for protein translation. Regulatory ncRNAs play an important role in regulating gene expression; for instance, microRNAs (miRNAs) and long-non-coding RNAs (lncRNAs) regulate gene expression at the translational and transcriptional levels, respectively. Figure 1 categorizes different types of regulatory ncRNAs based on sizes, structural features and biological functions (7,9–11).

Figure 1

Categorization of regulatory ncRNAs by structural characteristics and physiological functions.

Most ncRNA-omics studies of neurodegenerative diseases have focused on identifying expression changes of ncRNAs using transcriptomic approaches such as arrays or RNA sequencing technologies (12). However, little has been done to integrate ncRNA data with other types of omics data into mechanistic insights. This review aims to summarize the recent findings on the relevance and the roles of ncRNAs in AD from ncRNA-omics studies. Specifically, we summarize the recent findings on ncRNAs in AD including ncRNA signatures, interactions between ncRNAs and their target mRNAs, and ncRNA-regulated pathways in AD. We further review the potential of specific ncRNAs to serve as biomarkers and therapeutic targets for AD.

ncRNA-omics studies of AD

To curate the literature on ncRNA studies in AD, we searched the PubMed with keywords such as ‘microRNA, non-coding RNAs, cohort, omics, or Alzheimer’s disease’ and their combinations using the R package ‘RISmed’. Table 1 summarizes the ncRNA profiling of various tissues (e.g. brain, plasma, blood and cerebrospinal fluid [CSF]) from healthy subjects and patients with neurodegenerative diseases (13–34) as well as mouse models, and more details about these studies can be found in Supplementary Material, Table S1.

Table 1

Summary of ncRNA-omics studies of AD and other neurodegenerative diseases

Disease	Tissue source	N	Major technology	Diseases-associated ncRNAs	Reference
AD	Brain	96a	miRNA sequencing	miR-142a-5p, miR-146a-5p, miR-155-5p and miR-455-5p	14
		135a	miRNA profiling	miR-195	20
		114	Microarray	MIR7-3HG, AL109615.3, NEBL-AS1, ATP6V0E2-AS1, PDXDC2P-NPIPB14P, LOC441204, A2M-AS1, TGFB2-OT1 and LINC00672	17
		700	Nanostring	miR-132, miR-129-5p, miR-129-3p,miR-99b, linc-BRD9–1, linc-ADC, linc-RNFT2–1 and linc-CTSD-3	13
		1446	RobustRank aggregation	mir-17-5p, mir-106a-5p and mir-373-3p	19
	CSF	141	NGS	miR-27a-3p, miR-30a-5p, miR-34c, piR_019324, piR_019949 and piR_020364	21
		42	miRNA profile	miR-16-5p, miR-125b-5p, miR-451 and miR-605-5p	22
		89	miRNA profile	miR-222 and miR-125b	23
	Blood	185	TaqMan arrays	miR-27a-3p, miR-27b-3p and miR-324-5p	26
		27	NGS	miR-494-3p, miR-6894-3p, miR-421 and let-7a-3p	27
		19	TaqMan arrays	miR-342-5p	28
		807	miRNA profiling	The model was using miR-eQTLs	30
		71	miRNA profiling	miR-146b-5p and miR-15b-5p	31
		38	RT-qPCR	miR-144-5p, miR-221 and miR-374	32
	Plasma, brain, CSF	185a	Small RNA sequencing	miR-181a-5p, miR-146a-5p and miR-148a-3p	33
		1686	Database for annotation	27 miRNAs were differentially expressed in AD-related pathways, such as miR-7, miR-125b, miR-134, miR-137, miR-193b, miR-384 and miR-512	34
	Plasma and CSF	195	Microarray	miR-1273 g-3p	25
PD	Brain	65	RNA sequencing	LINC-PINT	16
DDS	Brain	300	Nanostring	miR-484	18
DLB	Brain	52	miRNA arrays	miR-133b, miR-34a, miR-7 and miR-137	15
FTD	CSF	89	Real-time PCR	miR-204-5p and miR-632	24
ALS	Blood	53	Real-time PCR	miR-192-5p, miR-192-3p, miR-1, miR-133a-3p, miR-133b, miR-144-5p, miR-19a-3p, miR-320c, miR-320a, let-7d-3p, miR-425-5p, miR-320b and miR-139-5p	29

AD, Alzheimer’s disease; FTD, frontotemporal dementia; ALS, amyotrophic lateral sclerosis; DDS, dementia and depressive symptoms; DLB, dementia with Lewy bodies; PD Parkinson’s disease; NGS, next generation sequencing.

a

These studies also used mouse samples.

ncRNA-omics of brain tissues in AD

A recent study of ncRNA-omics in AD performed a comprehensive analysis of the miRNA expression data in the brain from 28 independent studies with the support of a number of databases including TargetScan, Database for Annotation, Visualization and Discovery (DAVID), FunRich and String. A set of differentially expressed miRNAs between AD and normal control were identified and were found to be associated with amyloidogenesis, inflammation, tau phosphorylation, synaptogenesis, neurotropism, apoptosis and activation of cell cycle entry in AD (34). Among those differentially expressed miRNAs, miR-7, miR-9, miR-26, miR-34 and miR-125 were upregulated in AD brains (34). Moreover, an extensive ncRNA profiling was performed on the prefrontal cortex of AD subjects (miRNA profiling of 700 samples and lincRNA profiling of 540 samples (13)). The expression levels of miR-132, miR-129, linc-CTSD-3 and linc-BRD9-1 were found to be associated with brain amyloid pathology, while the levels of miR-132, miR-129, miR-99, linc-CTSD-3, linc-ADC and minc-RNFT2-1 were associated with brain neurofibrillary tangle (NFT) pathology. Leveraging the matched ncRNA-omics and transcriptomics data from the same set of individuals identified specific alterations in the cortical transcriptome of AD associated with changes in ncRNA expression (13). For example, the expression changes in miR-132 in AD were associated with the transcriptional changes of its putative target genes (e.g. EP300 and SIRT1) (13).

Similarly, miRNA-seq studies of 4- and 10-month-old AD transgenic mouse models of amyloidosis (APP^SWE/PS1^L166P) and tau pathology (THY-Tau22) revealed that six miRNAs (miR-10a-5p, miR-142a-5p, miR-146a-5p, miR-155-5p, miR-211-5p, and miR-455-5p) were upregulated in both AD mouse models. Among these microRNAs, four (miR-142a-5p, miR-146a-5p, miR-155-5p, and miR-455-5p) have altered expression levels in AD brains (33). However, the overexpression of these miRNAs in wild-type mice did not induce any cognitive impairment, suggesting either miRNA dysregulation may be reactive to AD pathology or miRNA dysregulation alone is not sufficient to induce AD-associated cognitive deficits.

The aforementioned studies mainly examined the correlation between ncRNAs and AD-related pathology. A recent study investigated the association between miRNA expression in the brain cortical region and late-life depression as a risk factor for dementia (18). It was found that the decreased expression levels of four miRNAs (miR-484, miR-26b-5p, miR-30d-5p and miR-197-3p) in the brain were associated with late-life depressive symptoms. Three of these miRNAs, including miR484 and miR-1973p, were associated with a faster decline in cognition. At the molecular level, miR484 was shown to modulate the expression of proteins involved in synaptic transmission and plasticity (18). More studies are needed to determine the role of ncRNAs in developing clinical symptoms of AD.

Together, these ncRNA-omics studies established the relevance of ncRNAs with AD and revealed ncRNA-modulated pathways other than amyloid and tau processes in AD pathogenesis. Future studies integrating ncRNA-omics with transcriptomics and proteomics in both human and AD mouse models will provide more insights into the functional impact of ncRNAs in AD.

Brain region-specific ncRNA-omics

Brain region-specific ncRNA changes were also investigated in neurodegenerative disorders. Abyadeh et al. employed a robust rank aggregation method and identified 1713 differentially expressed genes (DEGs) in five brain regions, including the hippocampus, the cerebellum, the entorhinal cortex, the frontal cortex and the temporal cortex. In all these brain regions except the cerebellum, most of the downregulated DEG signatures were enriched in retrograde endocannabinoid signaling and GABAergic synapse pathways which are potentially modulated by miR-17-5p, miR-106a-5p and miR-373-3p (19). It was previously reported that some members of the miR-17 and miR-106a families downregulate autophagy proteins (35,36), and miR-17 can regulate autophagy-mediated Aβ degradation in microglia (36).

Using Gene Expression Omnibus (GEO) database combined with supervised machine learning, expression patterns of long ncRNAs (lncRNAs) from the entorhinal cortex (EC), the hippocampus (HC), the post-central gyrus (PCG) and the superior frontal gyrus (SFG) were compared between AD subjects and healthy controls (17). It was found that about 146 lncRNAs were differentially expressed in the EC, HC and SFG brain regions between AD and healthy control brains, with 80% of these lncRNAs downregulated in the HC region, whereas all these lncRNAs were upregulated in the EC. For instance, miR7-3HG and MEG9 were the top two most downregulated lncRNAs in the HC, while VAC14-AS1 and LINC00472 were the top two most upregulated lncRNAs in the EC (17).

Another study examined miRNA expression patterns in patients with AD and dementia with Lewy bodies (DLB) using the Exiqon microarray technique. A hundred and fifty miRNAs were differently expressed between primary motor (MO) cortex and anterior cingulate (AC) cortex and they included miR-133b and miR-34a which were highly expressed in the MO region, and miR-137 and miR-7 which were upregulated in the AC region. The findings demonstrated the difference in brain regional vulnerability to DLB pathology (15). A previous study showed the neuroprotective role of miR-137 and miR-133b in attenuating Aβ-induced neurotoxicity (15,37). The difference in the changes of the expression levels of these miRNAs indicates brain region-specific vulnerability or resilience to the diseases.

The brain region-specific changes in ncRNA expression were explored in other neurodegenerative disorders as well. For example, lncRNA expression was measured in brain tissues of Parkinson’s disease (PD) patients and control donors by creating ribosomal-RNA-depleted RNA-Seq libraries followed by RNA Tag-Seq analyses (16). A total of 65 libraries were analyzed from the substantia nigra (SN), the amygdala and the medial temporal gyrus. The majority of the lncRNAs were commonly expressed in all the brain regions. Interestingly, 1,120 out of 3,718 lncRNAs identified were SN-specific with a subset of 13 lncRNAs differentially expressed between PD subjects and normal controls. An extensive validation study identified P53-induced non-coding transcript (LINC-PINT) as one key lncRNA differentially expressed in the SN-specific brain regions of PD patients compared to controls. LINC-PINT is a primarily neuronal transcript with functional roles implicated during the development and aging, as well as in AD and Huntington’s disease. Depletion of this lncRNA was associated with increased cell death induced by oxidative stress, suggesting a neuroprotective role of LINC-PINT in PD (16).

There are many limitations in current studies of brain region-specific ncRNA changes in AD and other neurodegenerative diseases. Some studies could not identify any ncRNA with brain-region specific changes (15,17,19). In addition, only a small number of differentially expressed ncRNAs were identified (16). Finally, the majority of the ncRNAs identified were highly expressed in neurons but were barely detectable in other brain cell types such as microglia and astrocytes. Therefore, single-cell sequencing of ncRNAs and mRNAs is needed to identify brain cell type specific changes in ncRNA expression and the correlation between ncRNAs and their targeted mRNAs.

Circulating ncRNA profiles in AD

The ncRNAs secreted into blood or CSF are named circulating ncRNAs. These molecules are extremely stable and can be divided into two categories: vesicle-associated and non-vesicle-associated (38). For their easy access, circulating ncRNAs have a big potential to serve as diagnostic and prognostic biomarkers of AD (Table 1).

Shigemizu et al. analyzed the miRNA profiles of blood samples from healthy control, mild cognitive impairments (MCI) and AD subjects using a network-based meta-analysis approach. It was reported that the expression levels of miR-5006-3p and miR-29a-3p and their targeted genes (SHC1 and PTEN) were altered with a lower SHC1 expression in the AD patients compared to the healthy subjects and a higher PTEN expression in AD compared to the MCI patients (30). Furthermore, the changes in the expression of the circulating miRNAs during disease progression were also investigated. Plasma miRNA levels were measured and compared between two groups of AD patients: fast decline in cognition (FDC) versus slow decline in cognition (SDC) based on changes in Mini-Mental State Exam scores within a span of 2 years. The patients with FDC were found to have lower plasma levels of miR-342-5p, which was predicted to target the genes involved in synaptic functions such as synaptotagmin, neurogranin and synaptojanin suggesting a potential role in predicting AD progression (28). Schneider et al. performed miRNA profiling of CSF exosomes derived from individuals with or at risk of developing frontotemporal dementia (FTD) using quantitative PCR arrays. It was found that miR-204-5p and miR-632 were highly expressed in pre-symptomatic individuals, suggesting potential roles as diagnostic biomarkers of FTD (24).

Expression changes in circulating ncRNAs could also help understand disease mechanisms. Using the blood samples from a set of MCI patients and healthy controls aged between 50–80 years, differential miRNA expression was investigated and further integrated with the matched epigenomic and lipidomic data (27). There were positive correlations between six miRNAs (miR-664, miR-432, let-7a-3p, miR-29a-3p, miR-421 and miR-450b-5p) and four fatty acids (FA (16:0), FA (18:0), FA (20:3), FA (20:4)), implicating a role of fatty acid impairment in AD pathogenesis (27). A study using a next-generation sequencing approach analyzed the exosomes derived from the CSF samples of the AD patients and the age-matched healthy controls. It was reported that a small ncRNA signature including three miRNAs (miR-27a-3p, miR-30a-5p, miR-34c) and three piRNAs (piR_019324, piR_019949, and piR_020364) were able to predict the conversion of MCI to AD with a high accuracy. The prediction accuracy was even higher when paired with Aβ₄₂/₄₀ ratio and tau phosphorylation. The expression of all these ncRNAs was increased in AD except for levels of piR_019324. The miRNA-targeted genes are associated with pathways involved in the HIF1α-related hypoxia as well as inflammatory processes via IGF1 and mTOR signaling (21).

In summary, the majority of circulating ncRNA studies examined the changes of ncRNA expression in human blood and CSF during disease progression, for the purpose of identification of effective diagnostic or prognostic biomarkers. Future studies are needed to better understand the roles of circulating ncRNAs in disease mechanisms such as linking peripheral and central disease processes, as well as facilitating cell–cell communication and/or disease spread.

Non-coding RNA-modulated regulatory networks

While a single miRNA usually targets a number of mRNAs, an individual mRNA can be regulated by multiple different miRNAs. The miRNA-mRNA network becomes even more complex with the participation of other ncRNAs, i.e. lncRNA (39). The crosstalk among coding and non-coding RNAs forms a complex regulatory network through miRNA response elements. According to the competing endogenous RNA (ceRNA) hypothesis (40), when the expression of a lncRNA is low, endogenous miRNA expression is upregulated to promote the degradation of its mRNA targets. In contrast, upregulation of lncRNA expression would downregulate miRNA expression with a subsequent increase in miRNA-targeted mRNAs (40).

A number of databases such as miRWalk (41), starBase (42) and LncBase (39,43) have been developed to include the known or predicted miRNA-mRNA and miRNA-lncRNA interactions. Moreover, experimentally validated AD-associated miRNA-target interactions can be obtained from databases such as miRTarBase (44). For example, a recent study generated lncRNA-miRNA-mRNA and coding-noncoding co-expression networks (39) and identified the target genes of four lncRNAs (i.e. MALAT1, OIP5-AS1, LINC00657 and lnc-NUMB-1), which include several genes such as such amyloid precursor protein (APP), presenilin-1 (PSEN1) and beta-secretase 1 (BACE1) that are involved in Aβ processes (39).

Another study investigated the ncRNA regulation of MAPT. As one type of ncRNAs, the MAPT-antisense RNA 1 (MAPT-AS1) is a type of natural antisense transcripts (NAT) of MAPT. Expression of MAPT-AS1 (as well as its minimal essential sequences) suppresses tau expression (45), whereas silenced MAPT-AS1 upregulates neuronal tau expression (45). These findings support the importance of ncRNAs as fine-tuning regulators of tau expression, which is tightly controlled at multiple levels such as NAT expression and enrichment of miRNA binding sites.

The regulatory networks of ncRNAs may be commonly shared among different neurodegenerative diseases. To identify common genes connected to AD and PD, Rana and collaborators initially performed an inner merge of GWAS (genome-wide association) reported gene loci for AD and PD. Interestingly, only one gene, HLA-DRB5, was reported for both diseases (46). Through literature mining and network analysis, miR-29 and miR-16 were found to target genes such as ELAVL1, SIRT1, PTGS2 and HLA-DRB5, and they form a core subnetwork shared by AD and PD (46). Both miRNAs had been previously reported to be differentially expressed in AD (22). This study suggests a potential commonly shared mechanism between AD and PD through specific ncRNAs-regulated networks.

Using the Omics datasets shown in Table 1, we performed a pairwise enrichment analysis of miRNAs derived from the Omics studies of human brain samples, as well as Gene Ontology (GO) pathways related to AD pathological processes including amyloid formation, tau pathology, protein-lipid metabolism, neuroinflammation—immune responses, endolysosomal pathway, synapse and myelin formation, as well as oxidative stress (Supplementary Material, Table S2). A link score was generated to predict the likelihood of a specific miRNA modulating AD pathways/networks by the –Log₁₀(FDR) with a cutoff value of 1.3 corresponding to FDR of 0.05 (Supplementary Material, Table S3). The impact of each miRNA identified from human brain studies on specific AD-related pathways is represented in Figure 2. For instance, it is predicted that miRNA-125 with a normalized link score of 27.35 could have a high impact on regulating the genes related to amyloid formation, while miRNA-132 is strongly associated with the genes involved in both amyloid formation and tau pathology pathways with the normalized link scores of 11.12 and 9.28, respectively. On the other hand, miRNA-99 has a stronger association with tau pathology than amyloid formation with the normalized link scores of 27.92 and 8.71, respectively. Several miRNA-modulated regulatory pathways identified from this approach were previously validated experimentally. For example, miR-195 was found to be lower in the brain with ApoE4^+/− MCI or early AD subjects. Consistent with this, elevation of miR-195 ameliorates cognitive deficits, amyloid plaque burden and tau hyperphosphorylation in the ApoE4^+/+ AD mouse models, supporting the regulatory roles of miR-195 in AD processes including amyloid formation, lipid metabolism, endolysosomal pathway and oxidative stress as summarized in Figure 2. This analysis provides a critical perspective of miR-195-modulated AD pathological processes.

Figure 2

Predicted miRNA-modulated regulatory pathways in AD. A pairwise enrichment analysis was performed with a list of miRNAs identified from human omics studies (Table 1) and AD-related pathways (Gene Ontology; Supplementary Material, Table S2). Only miRNA-pathway pairs with an FDR < 0.05 were retained, and a link score was generated as –log10(FDR), the higher the score the more relevant a miRNA to a pathway (i.e. GO term). The miRNA-pathway pairs were collapsed over the seven biological groups, and the highest link score was chosen to represent the relevance of a miRNA to the biological group as described in Supplementary Material, Table S3. Different AD-related pathways are presented in different color codes, and the font size is associated with normalized link scores generated from each pairwise enrichment analysis (miRNA, its target genes and GO pathways). The diagram was generated using word clouds website (Supplementary Material, Table S4).

On the other hand, several novel links between specific miRNAs and AD-related pathways have been identified (Figure 2). Future studies are needed to validate and characterize the functional relevance of predicted miRNA-modulated regulatory pathways/networks in AD pathogenesis. Nevertheless, this type of analysis highlights the power of applying advanced bioinformatic tools to study omics data and to better understand ncRNA-related disease mechanisms. Together, the complex networks among ncRNAs and protein-coding genes reveal the importance of multi-layer regulatory mechanisms in AD. It is speculated that dysfunction in any of these regulatory networks may lead to neurodegenerative processes.

Roles of ncRNAs as potential therapeutics for AD

Different from classical monotargeted therapies, a trend in developing therapeutics for treating complex polygenic disorders such as AD is to target multiple targets and pathways. miRNAs are particularly relevant in this regard (9,10) because of their pleiotropic and regulatory roles in disease processes (11,13,18,34,47). The miRNA-based therapies like synthetic antisense oligonucleotides (ASOs) (15,20) have been developed and tested for efficacy in rodents of AD models (9,19). These ASOs function either as miRNA mimics to downregulate expression of targeted genes or as anti-miR-mix (miRNA antagomirs) to inhibit the cognate miRNAs and thus upregulate target gene expression. For example, the antagomiRs targeting miR-181a-5p, miR-146a-5p and miR-148a-3p restored cognitive function in mouse models for dementia (19), whereas miR-132 mimic promoted adult hippocampal neurogenesis and rescued memory deficits in mouse model of AD (9). While no efficacious disease-modifying therapies are currently available for AD, miRNA-based therapeutics have great potential with progresses on ASO structural design and ASO delivery methods as well as mechanistic understanding of the miRNA-mRNA regulatory networks in AD (20).

Future perspectives

Advances in next-generation sequencing technologies, as well as increasingly available large multi-omics data, create an unprecedented opportunity for in-depth investigation of disease mechanisms and future development of AD therapeutics. At the same time, it poses a big challenge to develop and apply novel bioinformatics and statistical algorithms to integrate multi-layers of omics data such as genetic, genomic, transcriptomic, epitranscriptomic, proteomic, lipidomic and metabolomic data with ncRNA-omics data to unravel the mechanistic cascades across different layers of biological processes.

Single-cell RNA-seq studies have been widely applied to uncover cell-type specific processes and better understand molecular regulatory processes at the single cell level. However, single-cell ncRNA-seq studies lag behind, e.g. ncRNA profiles in a single cell remain largely unknown. There are different protocols for single-cell miRNA-seq being developed. A recent comprehensive comparison of these approaches demonstrated the feasibility of single-cell miRNA profiles as potential biomarkers for cancer patients (48). Furthermore, innovative approaches such as CsmiR have been developed to combine single-cell miRNA-mRNA co-sequencing data and putative miRNA-mRNA binding information to identify miRNA regulatory networks at the resolution of the single cell level (49). Future development of ncRNA profiling at the single cell level is urgently needed for studying neurodegenerative diseases.

In summary, here we review the most relevant ncRNA-omics studies of AD. Many questions await to be addressed, including the roles of ncRNAs in specific brain regions and specific brain cell types, mechanisms underlying regulatory processes of ncRNAs during AD pathogenesis, ncRNA changes specific to disease development and progression as diagnostic and prognostic biomarkers. In addition, the roles of ncRNAs such as lncRNAs, circular RNAs and enhancer RNAs in neurodegenerative diseases remain elusive. Future research should integrate multi-omics and multi-modal biological data into holistic biological network models of complex diseases that could guide investigation of new disease mechanisms, discovery of effective biomarkers and ultimately development of novel therapeutics.

Conflict of Interest statement. None declared.

Funding

National Institutes of Health RF1 (RF1AG048923 to D.C.), RF1 (RF1AG054014), RO1 (RO1AG068030), R56 (R56AG058655) and RF1 (RF1AG074010 to D.C. and B.Z.); Department of Veteran Affairs BLRD (I01BX003380); RR&D (I01RX002290 to D.C.); National Institutes of Health U01 (UO1AG046170), RF1 (RF1AG057440), RO1(RO1AG057907) and UO1 (UO1AG052411) to B.Z.

Authors’ Contributions

D.C. and B.Z. conceived the concept, outlined the study and revised the manuscript. E.W. and M.L.D. designed the analytic framework, performed the primary analyses and wrote the manuscript. L.E.R. participated the analyses and manuscript writing.

Author notes

Erming Wang and Mariana Lemos Duarte Co-first authors with equal contributions to the work.