Drug repurposing for Alzheimer's disease integrating transcriptome-wide association study and biological network analysis

Abstract

Background

The accumulation of particular protein deposits connected to molecular mechanisms is one of the many brain abnormalities associated with Alzheimer's disease (AD), a complex neurodegenerative illness. There are currently no effective disease-modifying treatments for AD.

Objective

This study attempts to identify potential AD therapeutics through a biological network-based drug repurposing strategy, focusing on drugs targeting important proteins and biological pathways involved in AD pathology.

Methods

A comprehensive biological network of AD-associated molecules and their transcription regulatory interactions is constructed. This computational approach integrates data from genome-wide association studies, multiple AD-related magnetic resonance imaging (MRI) derived phenotypes, biomolecular interactions, and gene expression profiles.

Results

The constructed AD sub-regulatory network reveals significant correlations between transcription factors showing changed gene expression in AD patients relative to controls. This strategy prioritizes drug candidates based on their mechanisms of action, reducing the risk of clinical trial failures and enhancing patient outcomes related to AD. A total of 43 drug candidates have been identified, including 28 FDA-approved drugs, 15 experimental and investigational drugs that may alter biological processes pertaining to important facets of AD pathology. Baricitinib and Gabapentin emerge as promising candidates for targeting AD-related biological processes in the cerebral cortex and hippocampus regions.

Conclusions

By combining biological network analysis and MRI-driven transcriptome-wide association study, this systematic drug repurposing strategy demonstrates promise for identifying novel therapeutic options for AD and offers potential implications for addressing other complex neurological disorders.

Keywords

Alzheimer's disease brain MRI drug repurposing image derived phenotype transcriptome-wide association study

Introduction

About 6.9 million Americans 65 and older suffer from Alzheimer's disease (AD), a progressive neurodegenerative illness that drastically lowers quality of life.^1,2 Cognitive decline and behavioral changes are hallmarks of AD. Drug development for AD, spanning various mechanistic classes, is known to be both expensive and time-consuming.³ Although clinical trials have shown that the FDA-approved medication Aducanumab is effective at binding aggregated amyloid-β (Aβ) and reducing Aβ deposition in clinical trials, its cost-effectiveness for AD treatment is debated.⁴ Drug repurposing which utilizes approved drugs for new indications, has emerged as a promising strategy for developing novel AD therapies. For example, calcium channel blockers and angiotensin receptor blockers have been investigated for treating AD because of their neuroprotective and anti-inflammatory qualities.⁵ Despite these initiatives, the majority of current drug repurposing strategies for AD are hypothesis-driven and depend on similarities between known pharmacological mechanisms and assumed pathogenic mechanisms.⁶ Nevertheless, this method frequently ignores the intricacy of AD pathology and might not take into consideration new information from objective, data-driven approaches. Furthermore, knowledge of AD-related biological processes remains limited, particularly regarding its pathological heterogeneity.⁷ Drug repurposing, which has potential applications in AD as well, offers a financially viable way to reduce the risks involved in drug research and development for complex neurological disorders like depression by combining various large-scale biological datasets.

Many national Alzheimer's genome projects have generated large amounts of genomic data rapidly, owing to the development of high-throughput technologies. Genome-wide association studies (GWAS) have been successfully used to identify new risk loci for AD.⁸ Recently, transcriptome-wide association studies (TWAS) have investigated the associations between genetically regulated gene expression and neurological disorders.⁹ TWAS studies have identified APOC1, CEACAM19 and TREM2 as genes associated with AD, which are implicated in immune response and synaptic function.¹⁰ Additionally, APOE4 expression levels and AD treatments were significantly correlated, highlighting the crucial role that lipid transport plays in the pathophysiology of AD.¹¹

In addition, brain magnetic resonance imaging (MRI) is widely used as a diagnostic tool for AD,¹² as it can detect structural and functional changes in the AD-affected brain.¹³ A large-scale epidemiological study UK Biobank integrates multimodal data including brain imaging, genetics and ongoing health outcomes.¹⁴ Brain image-derived phenotypes (IDPs) have been developed to quantify distinct aspects of brain structure and function via automated image processing pipelines based on brain MRI data.^15,16 These findings provide a valuable perspective on the genetic architecture of the brain, particularly in relation to neurological diseases and aging. For example, IDPs include measures such as the volume of white matter in specific brain regions, many of which are heritable according to GWAS of brain IDPs in the UK Biobank. Notably, some IDP-associated genes are involved in the transport of iron, which is related to neurodegenerative disorders.¹⁷ These results offer a variety of biomedical data supporting the integration of genetic and brain imaging phenotypes, potentially facilitating drug discovery for AD by targeting specific brain regions or pathways dysregulated by genetic risk factors.

Neuronal damage and Aβ accumulation are multifaceted factors linked to the development and progression of AD.¹⁸ As a result, therapeutic approaches that focus on multiple, interconnected pathways are increasingly recognized as crucial for optimizing AD treatment.¹⁹ Research on finding multi-target drug candidates is becoming more and more focused in line with this move towards multi-target therapeutics. Examples include medications that address the complex and multifaceted nature of AD by simultaneously targeting acetylcholinesterase (AChE) and N-methyl-D-aspartate (NMDA) receptors.²⁰ Recently, network medicine has emerged as a promising field for uncovering shared mechanisms and intrinsic phenotypes between AD and other diseases.²¹ Network-based drug repurposing has led to a paradigm shift in biomedical research with the vast data generated by high-throughput technologies. The emphasis has shifted from a “one target, one drug” approach to a “network target, multicomponent therapeutics” approach, highlighting a systematic perspective on illness management.^22,23 Network-based drug repurposing holds great potential as an innovative strategy for developing effective AD treatments.

In this study, we aim to integrate multiple types of biomedical data using a network analysis approach to identify effective drug candidates for treating AD. A novel computational strategy for network-based drug repurposing for AD is proposed based on transcriptome-wide association studies of brain IDPs and transcriptome data as well as drug-target interactions. The overall workflow of our strategy is depicted in Figure 1. First, we conduct a TWAS to investigate the association between AD-related brain IDPs and dysregulated genes based on the GWAS summary data from the UK Biobank brain MRI IDPs. Second, we prioritize repurposable drugs by assessing the proximity of their targets to AD-associated dysregulated proteins in a protein-protein interaction network. Third, we utilize AD-associated gene expression data to deduce transcriptional regulatory factors to identify drug candidates based on gene signature perturbation. Finally, consolidated evidence is sought from chemoinformatic resources to identify the most promising repurposable drug candidates for treating AD along with their potential mechanisms of action.

Figure 1.

Overall framework for the in silico identification of drugs for Alzheimer's disease through the transcriptome-wide association study (TWAS) and the network-based drug repurposing approach. The blue dashed box summarizes the TWAS procedures of AD-related IDPs. The orange dashed box shows drug repurposing based on AD-related IDPs’ dysregulated genes and biological network. The green dashed box summarizes drug-gene perturbation based on AD-related expression profiles. The black dashed box shows the results of repurposable drug candidates and mechanisms of action for the treatment of AD (Color figures available in online).

Methods

Selection of AD-related IDPs

This study incorporates 15 brain MRI volume-related IDPs as primary contributors identified from a larger set of 3935 based on their association with cognitive decline and AD pathology according to the previous study.²⁴ These selected brain MRI IDPs include total grey matter, white matter, whole brain, left hippocampus, right hippocampus, as well as segmented regions of the hippocampus (head, tail, and body) for both hemispheres.²⁵ By focusing on these specific IDPs associated with brain MRI volumes, this study aims to clarify their contributions to cognitive decline and their relevance to AD pathology. The targeted examination of these IDPs provides a promising strategy for understanding the molecular mechanisms underlying cognitive impairment and advancing potential AD treatments.

Transcriptome-wide association studies

For the selected IDPs, we extract GWAS summary data including approximately 40,000 participants from the UK Biobank Oxford Brain Imaging Genetics Server-BIG40 (www.open.win.ox.ac.uk/ukbiobank/big40/). We then conduct TWAS using the software FUSION. Specifically, to assess the association between genetically regulated gene expression and AD-related IDPs, we constructed Z-statistics as the test statistic for this association analysis:²⁶

Z = \frac{Z_{S G} \hat{β}}{\sqrt{{\hat{β}}^{'} V \hat{β}}}

(1)

where

Z_{S G}

represents a vector containing the Z-scores of cis-SNPs (single-nucleotide polymorphisms) from summary GWAS data,

\hat{β}

is the cis-SNPs’ effect size on gene expression. In this study, we utilize the pre-trained eQTL effect sizes from GTEx (Genotype-Tissue Expression Project) database in the cortex and hippocampus of brain tissues.

V

represents the covariance matrix of all cis-SNPs of a given gene g, which is estimated by using a population reference panel 1000 Genomes Project.

In accordance with the standard procedure for the FUSION approach,²⁷ only genes with significantly non-zero cis-acting heritability (cis-h²) are included in the analysis. Summary statistics are munged before analysis, i.e., SNPs are retained with an imputation INFO (an estimated quality measurement of imputation) greater than 0.9. Additionally, indels, strand ambiguous SNPs and SNPs with MAF (minor allele frequency) below 0.01 are excluded.²⁸ The Benjamini & Hochberg method is adopted to correct for multiple testing and control the false discovery rate. Since this study is more exploratory, a higher Type I error can be tolerated to achieve higher statistical power. Dysregulated genes are selected with a p-value less than 0.05, signifying statistical significance in the context of the analysis.

Drug-target-protein biological interactions

In our study, a comprehensive analysis of experimentally validated protein-protein interactions sourced from publicly available databases is compiled, comprising 327,924 interactions among 18,505 distinct proteins.²⁹ Additionally, we obtain data on 6361 drugs, including their mechanism of action, anatomical therapeutic chemical (ATC) code and clinical indication. These drugs encompass FDA-approved, investigational, experimental, and other categories. Furthermore, we extracted 22,605 drug-target interactions from the Drugbank database.³⁰ This dataset, detailing specific interactions between drugs and their target proteins, provides valuable a priori knowledge for understanding the mechanisms of drug action and potential therapeutic effects in AD. By integrating biological data on both protein-protein interactions and drug-target interactions, our study aims to uncover novel insights into drug mechanisms, potential targets, and therapeutic strategies for AD.

Network-based drug repurposing strategy

To identify potential drug candidates based on dysregulated genes associated with AD-related IDPs, this study employs a biomolecular network-based approach. This method involves constructing AD-associated biomolecular networks using the dysregulated genes and gene regulatory interactions related to AD-specific IDPs. Drug interventions are assessed based on the topological relationships of drug targets within the disease biomolecular network. The topological distances between dysregulated gene sets for AD-related IDPs and known drug target sets in a biomolecular network are computed. To evaluate the significance of the results, permutation tests are conducted by randomly selecting drugs from the biomolecular network nodes, matching the number of targets and topological properties of the network (Supplemental Figure 1). Corresponding drugs whose target sets are closer to the dysregulated gene sets of AD-related IDPs than would be expected by random chance are considered potential candidates. The distance $d (S, T)$ between TWAS gene set S and drug target set T, and z-statistics $z (S, T)$ are calculated as follows:

d (S, T) = \frac{1}{| T |} \sum_{t \in T} min_{s \in S} d (s, t)

(2)

z (S, T) = \frac{d (S, T) - μ_{d (S, T)}}{σ_{d (S, T)}}

(3)

where s represents the dysregulated gene in the TWAS gene set S, and t represents the protein target in the drug target set T.

Network separation calculation

Network separation is calculated by determining the topological distances between sets of biomolecules associated with different IDPs.²⁹ First, dysregulated genes associated with the IDPs are extracted from the biological network. Second, the biological network is analyzed using a network representation of edge lists. Thirdly, distance measures between the biomolecules in the network are calculated by identifying the shortest path between biomolecules and quantifying the separation between them using topological distance metrics, as demonstrated in the network-based drug repurposing strategy. Finally, network separation is calculated based on the distances between sets of biomolecules associated with distinct IDPs,

S_{A B} = ⟨ d_{A B} ⟩ \frac{⟨ d_{A A} ⟩ + ⟨ d_{B B} ⟩}{2}

(4)

where

d_{A B}

represents the average shortest path length between the biomolecule set of IDP A and the biomolecule set of IDP B. Note that

d_{A A}

and

d_{B B}

denote the average shortest path for nodes within the IDP-related biomolecule sets in the biological network.

S_{A B}

represents the degree of network separation between the biomolecule sets of IDP A and IDP B. Furthermore, the relationships between two IDP-gene modules and a drug-target module (drug-IDPs combination) are calculated based on the biological network. If S_AB < 0, two IDP-associated gene sets are located in the same network neighborhood, whereas S_AB ≥ 0 indicates topological separation between two IDP-associated gene sets.

AD transcriptome data acquisition

Transcriptomic data from AD patients are collected from the Gene Expression Omnibus (GEO) and include 125 individuals dying with varying severities of AD neuropathology. These data are obtained from the Mount Sinai/JJ Peters VA Medical Center Brain Bank (MSBB) AD cohort. The GSE84422 dataset contains differentially expressed genes across various brain regions, including the cortex and hippocampus from patients with varying degrees of AD-related pathological abnormalities in the MSBB cohort.³¹ A linear regression model is employed to correct covariates in the raw microarray data including sex, postmortem interval, pH and race. All data are available for research and have been deposited in the AMP-AD Knowledge Portal.³²

Major transcription factor analysis

A TF-centric regulatory network approach is used to explore AD-related mechanisms and potential preventive drugs by analyzing gene expression data. The identification of master regulator candidates is a crucial step in understanding the regulatory mechanisms underlying AD. The ChEA3 (ChIP-X Enrichment Analysis version 3) algorithm is applied out to assess gene-enriched cortical versus hippocampal regulators that are differentially expressed between AD cases and controls using differentially expressed genes in AD studies from the GSE84422 dataset.³³ The significance of the overlap between two gene sets is determined using Fisher's exact test with the background set of 20,000 genes by default. False discovery rates are calculated for each gene dataset individually using the Benjamini-Hochberg correction method. The master regulators are ranked in ascending order according to the mean rank of the ChEA3 algorithm, offering insights into key transcription factors involved in AD pathology. Using TFpertGEOup, TFpertGEOdn, ReMap, ENCODE, and ChIP-seq datasets, odds ratios are calculated for all TFs and TF-ChIP-seq pairs as follows,

O d d s r a t i o = \frac{U_{i n t e r s e c t i o n} / U_{o t h e r s}}{D_{i n t e r s e c t i o n} / D_{o t h e r s}}

(5)

where

U_{i n t e r s e c t i o n}

represents the intersection between upregulated genes and ChIP-seq targets,

D_{i n t e r s e c t i o n}

is the intersection between downregulated genes and ChIP-seq targets,

U_{o t h e r s}

is the set of upregulated genes but not ChIP-seq targets, and

D_{o t h e r s}

is the set of downregulated genes but not ChIP-seq targets.

Screening of drug candidates for the treatment of AD

To identify potential drug candidates for AD treatment, gene expression reverse perturbation analysis is conducted using AD-related DEGs. The largest repository of compound-induced gene expression profiles, LINCS, is utilized to investigate drug treatment-related gene expression reverse perturbation. LINCS provides a variety of available compounds and experimental conditions including cell lines, doses, and time of exposure, making it an ideal resource for drug-related gene expression reverse perturbation analysis. Specifically, we utilize the LINCS L1000 Chem Pert categories in combination with the Enrichr tool to screen for drug candidates.³⁴ The top 20 enriched TFs identified for each brain region in the major transcription factor analysis are ranked by an integrated mean rank score.³⁵ These TFs’ up- or down-regulated DEGs are selected for each AD case-control study and used as input in the Enrichr tool. A detailed summary of the TF-regulated DEGs from the AD patients dataset is provided in Supplemental Table 2. Drug candidates are screened based on their ability to significantly alter the expression of DEGs (p-value ≤ 0.05, under the Benjamini-Hochberg correction). This approach helps prioritize drugs that may reverse AD-related gene expression changes, offering promising therapeutic strategies.

Results

Identification of dysregulated genes associated with AD-related IDPs

Dysregulated genes associated with AD-related IDPs are identified using the FUSION software (see Methods). These IDPs represent quantitative measures of brain structure and function known to be altered in AD. Several noteworthy genes including CAP1, GSTP1 and GATAD2A, are recognized for each AD-related IDP. Dysregulated genes with their respective TWAS Z-scores for each IDP are shown in Figure 2. TWAS results reveal significant differences in dysregulated genes associated with IDPs of the cerebral cortex and hippocampus. For instance, ARL17A, a gene involved in intracellular protein transport and vesicle-mediated transport, exhibits significant dysregulation in the IDP of cerebral white matter. Moreover, the downregulation of microtubule-associated protein MAP1B in the right hippocampus volume is noted, which is a critical step in neurogenesis and rescues memory decline in AD. Biological functional enrichment analysis indicated that genes significantly associated with IDPs contribute to diverse nervous system biological processes. These include regulation of neuron maturation, positive regulation of synaptic plasticity and neuroepithelial cell differentiation (Supplemental Table 1).

Figure 2.

TWAS Z-scores of genes across AD-related IDPs. Significant genes are identified at p < 0.05 with Benjamini & Hochberg correction. Rows and columns are annotated by IDP name and dysregulated gene symbol, respectively. White spaces correspond to genes that are not significant in the TWAS. Blue shades indicate downregulation while red ones represent upregulation of gene expression (Color figures available in online).

Network-based drug repurposing for Alzheimer's disease

The distribution of AD-associated proteins in the protein-protein interactome is not arbitrary. Instead, they tend to group together to form local modules. Understanding how drugs work is essential to ensuring successful network-based drug repurposing. If a drug's targets establish a connection with the AD-associated module, it may be therapeutically effective. This connection is measured by a statistical z-score that captures the network proximity between a drug-target set (T) and an IDP-gene set (S), relying on the shortest distance between the drug target (t) and the disease protein (s) (see Methods). Our hypothesis is that exploring the network-based connections between drug targets and two IDP pairs, as well as the proteins associated with the IDPs, will help elucidate the mechanisms of action of drugs. Specifically, we utilize the separation measure s_AB to quantify the network proximity of IDP-associated gene sets A and B according to their gene localizations (see Methods).

From a biological network perspective, potentially effective drug-IDP combinations can be categorized into four topologically distinct classes: (a) Overlapping exposure occurs when two IDP-gene modules overlap with the drug-target module (Supplemental Figure 2A); (b) Indirect exposure occurs when one of the two overlapping IDP-gene modules intersects with the drug-target module (Supplemental Figure 2B); (c) Single exposure occurs when only one IDP-gene module, isolated from the other IDP-gene module, intersects with the drug-target module (Supplemental Figure 2C); (d) Complementary exposure occurs when each of the two distinct IDP-gene modules intersects with the drug-target module (Supplemental Figure 2D), respectively. Regarding the mechanisms of action of repurposable medications, these four classes may exhibit differences in clinical efficacy. To discern which of these drug-IDPs combinations exhibit mechanisms of action for the treatment of AD, FDA-approved drugs are selected.³⁶ For example, Tianeptine affects brain cortex-associated modules and indirectly regulates hippocampus-associated modules.

To validate the reliability of AD drug candidates identified through our biomolecular network-based drug repurposing strategy, we assessed various pharmacological and biological functional relationships. Notably, the cortex and hippocampus-associated IDP modules exhibited minimal overlap with each other (Figure 3(a)), suggesting distinct molecular signatures within these brain regions. Drugs with primary indications in oncology and dermatology are excluded based on the drug Anatomical Therapeutic Chemical (ATC) code classification. The mechanisms of action of repurposable drug-target modules and IDP-gene modules can be classified into four types: overlapping exposure (7.95%), indirect exposure (48.81%), single exposure (32.49%), and complementary exposure (0.75%) (Figure 3(b)). As depicted in Figure 3(c), the majority of drug candidates exhibit potential therapeutic effects for neurological diseases, including the ability to traverse the blood-brain barrier. Furthermore, most of the repurposable drug candidates with overlapping mechanisms of action are under the experimental category, while those with complementary exposure or indirect mechanisms are predominantly categorized as FDA-approved (Figure 3(d)).

Figure 3.

Statistical analysis of network-based drug repurposing for Alzheimer's disease. (a) Distribution of the separation scores between different IDP-associated gene modules. (b) The distinct categories of drug-IDP combinations and their proportions. (c) Number of different ATC code drugs in different drug mechanisms of action. (d) Proportion of registered drugs in the Drugbank database with different drug mechanisms of action.

Alzheimer's disease transcription factor analysis for drug perturbation

The elucidation of the mechanisms of drug repurposing for complex neurological diseases including AD, requires a comprehensive approach to analyzing the regulatory networks of high-throughput gene expression data. The differentially expressed genes associated AD are analyzed from the cerebral cortex and hippocampus of both healthy individuals and AD patients. Specifically, transcripts classified as transcription factors (TFs) in the Gene Ontology database with the DNA-binding transcription factor activity are analyzed. Based on the identified differentially expressed genes in AD, TFs enrichment analysis identifies the top 20 regulator candidates in the cerebral cortex and hippocampus, respectively. Our functional enrichment analysis reveals 23 key biological functions among differentially expressed genes in the cerebral cortex-associated region (as shown in Figure 4), including neuron migration, GABA receptor signaling, and frontal lobe dementia. In the hippocampus-associated region, 21 biological functions are significantly enriched among differentially expressed genes including long-term synaptic potentiation, neurotransmitter transmembrane transporter activity, and regulation of neuron differentiation. These findings highlight the significance of regulatory network analysis to identify potential drugs for the prevention and treatment of AD, particularly by elucidating region-specific molecular mechanisms underlying cognitive decline.

Figure 4.

Biological functions based on AD differentially expressed genes. Biological function enrichment analysis of differentially expressed genes in different brain regions of AD patients. Horizontal bar graph is gene expression enrichment significance in different biological functions. The red dashed line indicates the adjusted p-value at 0.05 (Color figures available in online).

The top 20 TFs are significantly altered in AD-related differentially expressed genes within the cerebral cortex and hippocampus regions (Supplemental Table 2). The most highly connected nodes in this regulatory network correspond to the MYT1L (Myelin Transcription Factor 1 Like) and PURG (Purine Rich Element Binding Protein G). The identification of differentially expressed genes targeted by AD-associated TFs is a critical step in elucidating the mechanisms of action involved in AD treatment. In this study, a total of 263 differentially expressed genes targeted by AD-associated TFs are identified in the cerebral cortex-related regions and 72 differentially expressed genes are identified in the hippocampus-related regions of AD patients (Supplemental Table 2). We use the Connectivity Map approach to identify up- or down-regulated differentially expressed genes in the regulatory network analysis that may be essential for treating AD in order to uncover possible therapeutic repurposing opportunities. Interestingly, differentially expressed genes associated with some IDPs include targets of approved psychotropic drugs such as Baricitinib and Loracarbef.

Repurposable drug candidates for Alzheimer's disease

The integration of biological network-based drug repurposing outcomes with transcriptomics, as illustrated in Figure 5(a), yields a comprehensive list of drug candidates exhibiting a robust connection with both IDP-associated genes and AD gene expression (as detailed in Supplemental Table 3). Among the identified drugs, 28 are FDA-approved, while 15 are experimental or investigational drugs. Notably, Figure 5(b) illustrates that the majority of the repurposable drug candidates regulate biological functions related to white matter, while those for the nervous system primarily affect functions associated with the brain hippocampus. Annotations from the Drugbank database, including the drug name, mechanism of action (MOA), and drug indication, correspond to these AD IDP-associated high-confidence genes as shown in Supplemental Table 4. Additionally, a manual query of the mechanism of action of the candidate drugs results in Table 1, presenting a list of potential FDA-approved drugs, their indications, and their relevance to AD research.

Figure 5.

Repurposable drugs on the AD-related IDPs. (a) The enrichment score (ES) from the biological functional enrichment analysis of AD gene expression data. The p-values are calculated between drug targets and IDPs-related genes by the network-based drug repurposing approach. (b) Occurrence of different ATC categorical drugs for AD-related IDPs.

Table 1.

Drug candidates for the treatment of AD based on IDPs-associated genes and transcriptome datasets.

Drug name	Targets/MOA	Indication	AD research progress	Potential MOA
Thiethylperazine	DRD2, DRD1, DRD4	Nausea and vomiting	Alzheimer's Treatment/ Clinical investigation	The Amyloid Beta draining effect
Baricitinib	JAK1, JAK2, PTK2B, JAK3, CYP3A4, SLC22A8, ABCG2, SLC47A2, SLC22A6, SLC22A2, SLCO1B3	Rheumatoid arthritis	Alzheimer's Treatment/ Clinical investigation	Regulation of neurotransmitters
Apremilast	PDE4	Treatment of inflammatory conditions	Alzheimer's Treatment/ Clinical investigation	Neuroprotective and senolytic properties
Levodopa	DRD1, DRD5, DRD2, DRD3, DRD4	Parkinson's disease	Ameliorates learning and memory deficits in AD/Experiments	Ameliorating learning and memory deficits
Gabapentin	CACNA2D1, CACNA2D2, CACNA1B, ADORA1, GRIN1, GABBR1, GABBR2, BCAT1, CYP3A4	Antiepileptic drugs	Improve behavior disturbance in AD	/
Zimelidine	SLC6A4, MAOB, MAOA, CYP3A4, ABCB1	Depression	Cognitive measures in AD/ Clinical investigation	A serotonin reuptake blocker
Methylphenidate	SLC6A3, SLC6A2, SLC6A4	Narcolepsy	Alzheimer's Treatment/ Clinical investigation	/

A dopamine precursor called Levodopa is mostly used to treat Parkinson's disease. In a mouse model, Levodopa has been shown to improve learning and memory deficits, suggesting that it may be a promising medication for AD.^37,38 Norgestrel, a potent progestin, exhibits direct neuroprotective effects, suggesting its potential therapeutic value in neurological conditions such as AD. Methylphenidate may help improve memory and attention because it is known to raise dopamine levels in the brain. It is clinically reported to improve global cognition with minimal adverse events in AD.³⁹ Preliminary low-grade evidence based on case series indicates that gabapentin could be beneficial for behavioral and psychological symptoms of dementia in patients with AD.⁴⁰ Flumazenil is a benzodiazepine receptor agtagonist, which affects memory retrieval in mice.⁴¹ Apremilast is shown to have immunomodulator, neuroprotective and senolytic properties. Therefore, Apremilast, like other PDE4, inhibitors may help alleviate symptoms in AD patients.⁴²

The relationship between drug mechanisms of action and IDP-associated genes is not always direct, as outlined in Supplemental Table 5. The neurotransmitter serotonin, the G protein-coupled receptor family, opioid receptors, cholinergic receptors, and other gene classes could be used to broadly classify these drug targets. For instance, increased methylation levels of opioid receptor κ1 (OPRK1) and opioid receptor δ1 (OPRD1) are associated with AD, indicating that they may be used as methylation biomarkers for the diagnosis of AD.⁴³ Wang et al. identified albumin (ALB) as strongly associated with iron death in AD, and ALB could be a potential iron death-related biomarker for AD diagnostic and therapeutic monitoring.⁴⁴ ABC transporter proteins are implicated in the regulation of Aβ levels in the brain, and intracerebral accumulation of Aβ peptides is a crucial process in AD. The transporter protein ABCC1 has been shown by Krohn et al. to play a significant role in the clearance and accumulation of Aβ in the brain.⁴⁵ In addition, researchers have found that MAO-B activity increases with age in brain tissue, cerebrospinal fluid, and platelets in AD patients. This suggests that MAO-B inhibitor drugs could be effective in treating AD.⁴⁶ Furthermore, potential therapeutic targets based on the mechanism of oxidative stress in AD are identified through constructing pharmacological networks. It is found that the progesterone receptor and estrogen receptor 1 may be closely related to neuroprotective effects and pathogenesis. Human and animal studies have shown that adenosine receptors (ARs) activity enhances cognition and neuroprotection, potentially improving cognitive impairment in animal models of AD.⁴⁷

Discussion

Network medicine provides a powerful approach for exploring the complex mechanisms of biological systems in AD and predicting drug-disease interactions.⁴⁸ Consequently, network-based drug repurposing strategies have emerged as a potential method to identify existing drugs for new indications.⁴⁹ Numerous studies have used network approaches to identify AD-related genes and pathways as potential drug targets. One such approach leverages known AD risk genes and analyzes drug-induced gene expression changes within the human protein-protein interaction network, effectively integrating genetic risk with drug perturbation data.⁵⁰ For instance, VCAM1 has been identified as a potential drug target gene involved in the blood-brain barrier and immune cell transmigration, particularly in the parahippocampal gyrus.⁵¹ In another study, Siavelis et al. utilized bioinformatics analysis and gene signatures to uncover potential anti-AD drugs.¹⁹ They curated potential anti-AD agents, which were further scrutinized for molecular similarity, pathway/gene ontology enrichment and network analysis. Similarly, protein kinase C (PKC) pathway has been recognized as a potential therapeutic target for AD through biological network analysis.⁵² Zhou et al. demonstrated that decreased PKC activity in AD patients’ brains. The activation of PKC with bryostatin-1 showed promise in enhancing cognitive function in a mouse model of AD. While GWAS primarily identifies genetic loci associated with AD, network approaches can help prioritize and functionally interpret these findings. For example, network analyses have highlighted the importance of GSK3β and CDK5, two key proteins involved in AD pathology that are often found to be dysregulated in GWAS studies.⁵³ In addition to these established pathways, emerging research highlighted how ER stress and the unfolded protein response contribute to neuronal dysfunction and suggesting that targeting these pathways could offer novel therapeutic opportunities.⁵⁴ Furthermore, Sharma et al. explored the mechanistic pathways through which chronic stress induces gut dysbiosis, which in turn exacerbates the neuropathological cascade of AD, highlighting the stress–dysbiosis axis as a potential therapeutic target.⁵⁵

AD may share common pathological mechanisms or intrinsic phenotypes with other illnesses. Consequently, drugs originally developed for the treatment of metabolic disorders, mental illnesses, and other clinical indications could potentially benefit AD patients. For instance, Thiethylperazine, a drug initially indicated for nausea and vomiting, has been shown to activate transport protein ABCC1 to aid in the removal Aβ.⁵⁶ Adding to the repertoire of potential therapeutic agents, research on resveratrol, a natural polyphenol, has demonstrated its neuroprotective effects in AD by promoting non-amyloidogenic processing of amyloid precursor protein and enhancing Aβ clearance.⁵⁷ Several repurposed drugs were collected for AD through the PRISMA 2020 systematic review and Agarwal et al. found Baricitinib and Lonafarnib for AD treatment consistent with ours.^58,59 Furthermore, we also identified other repurposed drugs reported such as Mepivacaine and Stavudine in clinical investigations. Moreover, drug repurposing approaches have proven valuable in identifying novel AD drug targets. For instance, farnesyltransferase (FNTA) and its downstream signaling are upregulated in postmortem AD brains, which can be a therapeutic target in treating AD.⁶⁰ In addition, enhanced memory is shown in the rat model of AD via inhibition of the JAK2/STAT3 pathway.⁶¹ In total, these findings suggest the potential of network-guided drug repurposing as a powerful and efficient strategy for accelerating the development of novel AD therapeutic regimens.

Our network-based strategy provides a valuable way to identify possible AD drug candidates. To improve the reliability of drug repurposing for the treatment of AD, several limitations and challenges need to be resolved. First, it is difficult to fully represent AD in our network model due to its intrinsic complexity as a polygenic disease impacted by multiple factors. Second, the reliability of identified drug-disease relationships is subject to the limitations of sample size. Furthermore, drug repurposing can also be challenging since the identified drug-disease relationships require further experimental validation. Larger datasets and rigorous validation procedures are needed to overcome these challenges and realize the therapeutic potential of our findings.

Conclusion

We have effectively integrated TWAS and a network-based drug repurposing approach to investigate the changes in gene expression within transcriptional profiles associated with drug effects. By integrating TWAS of AD-related IDPs and TF regulatory units, we have identified 28 FDA-approved drug candidates with potential in reversing AD-related phenotypes. In summary, our study presents a novel network-based drug repurposing strategy to predict drug candidates for treating AD patients. This innovative approach utilizes network analyses to tap into multiple types of biomedical data, offering a promising avenue for identifying potential therapeutic strategies for AD and other complex neurological disorders. Multiple types of biomedical data are leveraged by biological network analysis to identify effective drug candidates for AD. This comprehensive approach holds great potential in advancing the field of drug repurposing and accelerating the discovery of effective treatments for neurodegenerative conditions like AD.

Footnotes

Acknowledgments

The authors have no acknowledgments to report.

ORCID iDs

Xin Wang

Yan Dora Zhang

Author contributions

Xin Wang: Conceptualization, Data curation, Formal analysis, Methodology, Resources, Software, Validation, Writing – original draft, Writing – review & editing.

Meng Wang: Data curation, Writing – original draft, Writing – review & editing.

Han Wang: Data curation, Formal analysis, Writing – original draft, Writing – review & editing.

Guosheng Yin: Conceptualization, Writing – original draft, Writing – review & editing.

Yan Dora Zhang: Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Writing – original draft, Writing – review & editing.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by Hong Kong Research Grants Council General Research Fund [17307324].

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability statement

The data used in the manuscript are compiled from several public biomedical data sources including: DrugBank (version: 5.1.9, https://go.drugbank.com/), UK Biobank Brain imaging-derived phenotypes GWAS Summary statistics (https://open.win.ox.ac.uk/ukbiobank/big40/), GEO (www.ncbi.nlm.nih.gov/geo/ under the accession numbers GSE84422), Gene Ontology (http://geneontology.org/), ChEA3 (https://maayanlab.cloud/chea3/), LINCS (https://clue.io/), MeSH (www.ncbi.nlm.nih.gov/mesh/), protein-protein interactome. TWAS analysis is carried out using the FUSION (), Drug repurposing analyses and figures are performed using Python 3.7 and R 4.1.2.

Supplemental material

Supplemental material for this article is available online.

References

Alzheimer's Association. 2024 Alzheimer's disease facts and figures. Alzheimers Dement 2024; 20: 3708–3821.

Lopez-Bueno

Yang

Stamatakis

, et al. Moderate and vigorous leisure time physical activity in older adults and Alzheimer's disease-related mortality in the USA: a dose-response, population-based study. Lancet Healthy Longev 2023; 4: e703–e710.

Kodamullil

Zekri

Sood

, et al. Trial watch: tracing investment in drug development for Alzheimer disease. Nat Rev Drug Discov 2017; 16: 819.

Sinha

Barocas

. Cost-effectiveness of aducanumab to prevent Alzheimer's disease progression at current list price. Alzheimers Dement (N Y) 2022; 8: e12256.

Ballard

Aarsland

Cummings

, et al. Drug repositioning and repurposing for Alzheimer disease. Nat Rev Neurol 2020; 16: 661–673.

Rodriguez

Hug

Todorov

, et al. Machine learning identifies candidates for drug repurposing in Alzheimer's disease. Nat Commun 2021; 12: 1033.

Boyle

Leurgans

, et al. Attributable risk of Alzheimer's dementia attributed to age-related neuropathologies. Ann Neurol 2019; 85: 114–124.

Wightman

Jansen

Savage

, et al. A genome-wide association study with 1,126,563 individuals identifies new risk loci for Alzheimer's disease. Nat Genet 2021; 53: 1276–1282.

Fabbri

Pain

Hagenaars

, et al. Transcriptome-wide association study of treatment-resistant depression and depression subtypes for drug repurposing. Neuropsychopharmacology 2021; 46: 1821–1829.

10.

Gockley

Montgomery

Poehlman

, et al. Multi-tissue neocortical transcriptome-wide association study implicates 8 genes across 6 genomic loci in Alzheimer's disease. Genome Med 2021; 13: 76.

11.

Uddin

Kabir

Al Mamun

, et al. APOE And Alzheimer's disease: evidence mounts that targeting APOE4 may combat Alzheimer's pathogenesis. Mol Neurobiol 2019; 56: 2450–2465.

12.

Frisoni

Fox

Jack

Jr , et al. The clinical use of structural MRI in Alzheimer disease. Nat Rev Neurol 2010; 6: 67–77.

13.

Schilling

Archer

Yeh

, et al. Aging and white matter microstructure and macrostructure: a longitudinal multi-site diffusion MRI study of 1218 participants. Brain Struct Funct 2022; 227: 2111–2125.

14.

Sudlow

Gallacher

Allen

, et al. UK Biobank: an open access resource for identifying the causes of a wide range of complex diseases of middle and old age. PLoS Med 2015; 12: e1001779.

15.

Miller

Alfaro-Almagro

Bangerter

, et al. Multimodal population brain imaging in the UK Biobank prospective epidemiological study. Nat Neurosci 2016; 19: 1523–1536.

16.

Alfaro-Almagro

Jenkinson

Bangerter

, et al. Image processing and quality control for the first 10,000 brain imaging datasets from UK Biobank. Neuroimage 2018; 166: 400–424.

17.

Elliott

Sharp

Alfaro-Almagro

, et al. Genome-wide association studies of brain imaging phenotypes in UK Biobank. Nature 2018; 562: 210–216.

18.

Chew

Solomon

Fonteh

. Involvement of lipids in Alzheimer's disease pathology and potential therapies. Front Physiol 2020; 11: 598.

19.

Siavelis

Bourdakou

Athanasiadis

, et al. Bioinformatics methods in drug repurposing for Alzheimer's disease. Brief Bioinform 2016; 17: 322–335.

20.

Uddin

Al Mamun

Kabir

, et al. Multi-target drug candidates for multifactorial Alzheimer's disease: aChE and NMDAR as molecular targets. Mol Neurobiol 2021; 58: 281–303.

21.

Swarup

Hinz

Rexach

, et al. Identification of evolutionarily conserved gene networks mediating neurodegenerative dementia. Nat Med 2019; 25: 152–164.

22.

Hopkins

. Network pharmacology: the next paradigm in drug discovery. Nat Chem Biol 2008; 4: 682–690.

23.

Wang

Zheng

, et al. TCM Network pharmacology: a new trend towards combining computational, experimental and clinical approaches. Chin J Nat Med 2021; 19: 1–11.

24.

Tank

Ward

Flegal

, et al. Association between polygenic risk for Alzheimer's disease, brain structure and cognitive abilities in UK biobank. Neuropsychopharmacology 2022; 47: 564–569.

25.

Congdon

Sigurdsson

. Tau-targeting therapies for Alzheimer disease. Nat Rev Neurol 2018; 14: 399–415.

26.

Hao

Wang

Zhang

, et al. Prediction of Alzheimer's disease-associated genes by integration of GWAS summary data and expression data. Front Genet 2018; 9: 653.

27.

Gusev

Shi

, et al. Integrative approaches for large-scale transcriptome-wide association studies. Nat Genet 2016; 48: 245–252.

28.

Adams

Reay

Cairns

. Multiomic prioritisation of risk genes for anorexia nervosa. Psychol Med 2023; 53: 6754–6762.

29.

Menche

Sharma

Kitsak

, et al. Disease networks. Uncovering disease-disease relationships through the incomplete interactome. Science 2015; 347: 1257601.

30.

Wishart

Feunang

Guo

, et al. Drugbank 5.0: a major update to the DrugBank database for 2018. Nucleic Acids Res 2018; 46: D1074–D1082.

31.

Wang

Roussos

McKenzie

, et al. Integrative network analysis of nineteen brain regions identifies molecular signatures and networks underlying selective regional vulnerability to Alzheimer's disease. Genome Med 2016; 8: 104.

32.

Greenwood

Montgomery

Kauer

, et al. The AD knowledge portal: a repository for multi-omic data on Alzheimer's disease and aging. Curr Protoc Hum Genet 2020; 108: e105.

33.

Keenan

Torre

Lachmann

, et al. ChEA3: transcription factor enrichment analysis by orthogonal omics integration. Nucleic Acids Res 2019; 47: W212–W224.

34.

Chen

Tan

Kou

, et al. Enrichr: interactive and collaborative HTML5 gene list enrichment analysis tool. BMC Bioinformatics 2013; 14: 128.

35.

Swahn

Olmer

Lotz

. RNA-binding proteins that are highly expressed and enriched in healthy cartilage but suppressed in osteoarthritis. Front Cell Dev Biol 2023; 11: 1208315.

36.

Rivero-Garcia

Castresana-Aguirre

Guglielmo

, et al. Drug repurposing improves disease targeting 11-fold and can be augmented by network module targeting, applied to COVID-19. Sci Rep 2021; 11: 20687.

37.

Ambree

Richter

Sachser

, et al. Levodopa ameliorates learning and memory deficits in a murine model of Alzheimer's disease. Neurobiol Aging 2009; 30: 1192–1204.

38.

Soeda

Yoshikawa

Almeida

, et al. Toxic tau oligomer formation blocked by capping of cysteine residues with 1,2-dihydroxybenzene groups. Nat Commun 2015; 6: 10216.

39.

Rosenberg

Lanctot

Drye

, et al. Safety and efficacy of methylphenidate for apathy in Alzheimer's disease: a randomized, placebo-controlled trial. J Clin Psychiatry 2013; 74: 810–816.

40.

Supasitthumrong

Bolea-Alamanac

Asmer

, et al. Gabapentin and pregabalin to treat aggressivity in dementia: a systematic review and illustrative case report. Br J Clin Pharmacol 2019; 85: 690–703.

41.

Raut

Bolegave

Marathe

. Effect of flumazenil on memory retrieval determined by trial-to-criteria inhibitory avoidance method in mice. J Pharmacol Pharmacother 2016; 7: 24–26.

42.

Gomaa

Farghaly

HSM

Ahmed

, et al. Intermittent treatment with Apremilast, a phosphodiesterase-4 inhibitor, ameliorates Alzheimer's-like pathology and symptoms through multiple targeting actions in aged T2D rats. Int Immunopharmacol 2023; 117: 109927.

43.

Liu

, et al. Elevated methylation of OPRM1 and OPRL1 genes in Alzheimer's disease. Mol Med Rep 2018; 18: 4297–4302.

44.

Wang

Chen

Shao

. Identification of ferroptosis-related genes in Alzheimer's disease based on bioinformatic analysis. Front Neurosci 2022; 16: 823741.

45.

Krohn

Lange

Hofrichter

, et al. Cerebral amyloid-beta proteostasis is regulated by the membrane transport protein ABCC1 in mice. J Clin Invest 2011; 121: 3924–3931.

46.

Ozdemir

Alagoz

Bahcecioglu

, et al. Monoamine oxidase-B (MAO-B) inhibitors in the treatment of Alzheimer's and Parkinson's disease. Curr Med Chem 2021; 28: 6045–6065.

47.

Chen

. Adenosine receptor control of cognition in normal and disease. Int Rev Neurobiol 2014; 119: 257–307.

48.

Greene

Loscalzo

. Putting the patient back together – social medicine, network medicine, and the limits of reductionism. N Engl J Med 2017; 377: 2493–2499.

49.

Cheng

Desai

Handy

, et al. Network-based approach to prediction and population-based validation of in silico drug repurposing. Nat Commun 2018; 9: 2691.

50.

Peng

Yuan

Xin

, et al. Screening novel drug candidates for Alzheimer's disease by an integrated network and transcriptome analysis. Bioinformatics 2020; 36: 4626–4632.

51.

Liu

Manuel

Dai

, et al. Identifying candidate genes and drug targets for Alzheimer's disease by an integrative network approach using genetic and brain region-specific proteomic data. Hum Mol Genet 2022; 31: 3341–3354.

52.

Zhou

Chen

Zhong

, et al. Molecular identification of protein kinase C beta in Alzheimer's disease. Aging (Albany NY) 2020; 12: 21798–21808.

53.

Fang

Zhang

Wang

, et al. Artificial intelligence framework identifies candidate targets for drug repurposing in Alzheimer's disease. Alzheimers Res Ther 2022; 14: 7.

54.

Uddin

Tewari

Sharma

, et al. Molecular mechanisms of ER stress and UPR in the pathogenesis of Alzheimer's disease. Mol Neurobiol 2020; 57: 2902–2919.

55.

Sharma

Singh

Garg

, et al.

Dysbiosis and Alzheimer's disease: a role for chronic stress?

Biomolecules 2021; 11: 678.

56.

Cummings

Lee

Zhong

, et al. Alzheimer's disease drug development pipeline: 2021. Alzheimers Dement (N Y) 2021; 7: e12179.

57.

Rahman

Akter

Bhattacharya

, et al. Resveratrol and neuroprotection: impact and its therapeutic potential in Alzheimer's disease. Front Pharmacol 2020; 11: 619024.

58.

Agarwal

Katare

Jakhmola-Mani

. Foresee novel targets for Alzheimer's disease by investigating repurposed drugs. CNS Neurol Disord Drug Targets 2022; 22: 1209–1231.

59.

Page

McKenzie

Bossuyt

, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. Br Med J 2021; 372: n71.

60.

Cuddy

Alia

Salvo

, et al. Farnesyltransferase inhibitor LNK-754 attenuates axonal dystrophy and reduces amyloid pathology in mice. Mol Neurodegener 2022; 17: 54.

61.

Hindam

Ahmed

El Sayed

, et al. Repositioning of baricitinib for management of memory impairment in ovariectomized/D-galactose treated rats: a potential role of JAK2/STAT3-PI3K/AKT/mTOR signaling pathway. Life Sci 2024; 351: 122838.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

1.14 MB

0.02 MB

0.03 MB

0.01 MB