Bibliometric analysis of pathological mechanisms in Alzheimer's disease: Applications based on mouse models

Abstract

Background

Alzheimer's disease (AD) is a progressive neurodegenerative disorder marked by memory loss and cognitive decline. Animal models play a key role in exploring its pathophysiological mechanisms.

Objective

To analyze global research trends and knowledge structure in AD pathophysiological mechanisms based on animal models.

Methods

Publications from 2014 to 2023 were retrieved from the Web of Science Core Collection. CiteSpace and VOSviewer were used for bibliometric analysis and data visualization.

Results

A total of 2169 publications were identified, with a steady growth trend. The United States and China were the leading contributors, with Harvard University as a major collaborative hub. The Journal of Alzheimer's Disease published the most articles, while the Journal of Neuroscience had the highest co-citation frequency. Holtzman DM was a key author in the field. Nine keyword clusters were identified, including insulin resistance, amyloid beta, and oxidative stress. Emerging topics include synapse loss, gut microbiota, and NLRP3 inflammasome.

Conclusions

This study provides a concise overview of global research on AD pathophysiological mechanisms in animal models, offering valuable insights for future research directions.

Introduction

Alzheimer's disease (AD) is a neurodegenerative disorder characterized by progressive cognitive impairment and memory decline, marked by the gradual and irreversible deterioration of the brain, leading to cognitive dysfunction and memory loss.^1,2 The prevalence of AD increases significantly with age, particularly in individuals aged 65 and older. ³ Globally, AD represents the leading cause of dementia, contributing to an estimated 60% to 80% of all diagnosed cases. ⁴ Projections indicate that by 2050, over 100 million people worldwide will be living with AD.^5,6 The pathophysiology of AD involves a range of complex biological processes, including amyloid-β (Aβ) and tau protein pathology, neuroinflammation, oxidative stress, and genetic factors.^7–10 While significant progress has been made in understanding these mechanisms, the precise interactions among them and the initiating factors of AD remain incompletely elucidated.

Animal models play a critical role in studying AD pathogenesis and developing therapeutic strategies, particularly for early diagnosis and treatment evaluation. Mice are the most commonly used model in AD research, not only due to their ease of genetic manipulation, but also because a wide array of genetically modified mouse strains have been developed to model specific pathological mechanisms of AD. ¹¹ While mice are not ideal for modeling complex human-like learning and memory processes—as rats generally perform better on sophisticated operant tasks—they can still be tested on relatively straightforward and widely adopted tasks, such as contextual fear conditioning and Morris water maze tests. These protocols are simple to implement and facilitate cross-laboratory comparisons.^12–14 However, current AD mouse models are primarily based on early-onset familial AD, which makes it difficult to fully replicate the complex etiology and pathology of late-onset sporadic AD. ¹⁵ Future research should prioritize the development of innovative models that better emulate the early pathology, cognitive dysfunction, and environmental influences associated with AD.

Given the significant global health burden posed by AD, there has been a marked increase in research efforts worldwide, resulting in a substantial growth in relevant publications. Many scholars have conducted extensive research on the pathological mechanisms of AD, treatment strategies, and the application of AD mouse models. However, despite the large body of literature, traditional systematic review methods often struggle to comprehensively and systematically summarize the research progress in this field.

Bibliometrics, an interdisciplinary field employing quantitative methods to analyze the knowledge system, provides a robust approach for evaluating and synthesizing scholarly output. ¹⁶ Bibliometrics allows for qualitative and quantitative analysis of the contributions and collaborations of authors, institutions, countries, and journals, and can also evaluate the development and emerging trends in scientific retrieval.^17–19 Compared to systematic reviews and meta-analyses, bibliometric analysis offers a more detailed, timely, and objective framework to visualize and assess research landscapes, monitor field evolution, and predict future developments. ²⁰ Therefore, bibliometrics is highly suitable for the evaluation and research of AD mouse models and their pathophysiological mechanisms.

This study employs two widely used bibliometric tools, CiteSpace and VOSviewer, to perform qualitative and quantitative analyses of the literature published between 2014 and 2024 on AD mouse models and their pathophysiological mechanisms. By doing so, this study aims to objectively map the knowledge domain and highlight emerging trends in this critical area of research.

Methods

This study is based on the Web of Science Core Collection (WOSCC). Raw data were downloaded and extracted from the Science Citation IndexExpanded (SCI-E) database. The detailed search was performed using the following criteria:TS = ((“Alzheimer's Disease” OR “Alzheimer disease”) AND (“mouse model” OR “murine model” OR “transgenic mice”) AND (“pathological mechanism” OR “pathology” OR “disease mechanism”)), The publication language was limited to English, and the retrieval time span was set from 2014 to 2023. Document types were restricted to Articles and Reviews. All publications were searched and downloaded on the same day. Additionally, selected papers were screened to ensure their relevance to the study. Since the data used in this research were sourced from the publicly accessible WOSCC database and did not involve human participants, informed consent or ethical approval was not required. A flowchart illustrating the literature search and selection process is shown in Figure 1.

Figure 1.

Flowchart of the literature search and screening process.

Following a thorough literature screening, we exported the records in ‘Plain Text File’ format, including both the ‘Full Record’ and ‘Cited References’, and due to the export limit of 500 records per batch in the Web of Science platform, we divided the data into multiple files named “download_1_500,” “download_501_1000,” “download_1001_1500,” “download_1501_2000,” and “download_2001_2169.”The files were then imported into the Bibliometrix R package, CiteSpace (version 6.3.R3), and VOSviewer (version 1.6.20) for bibliometric analysis. Subsequently, we conducted an extensive bibliometric analysis, focusing on key characteristics, such as countries/regions, institutions, authors, journals, co-cited references, and keywords with notable citation bursts.

CiteSpace, developed by Professor Chaomei Chen, is a bibliometric visualization tool designed specifically to detect emerging trends and significant research areas in scientific literature.^21,22 VOSviewer, developed by Leiden University, is a software tool that enables the creation of visual bibliometric maps. ²³ Bibliometrix, developed in the R environment, is an open-source tool designed for comprehensive science mapping analysis. ²⁴ It provides a suite of quantitative research methods, and with its web interface Biblioshiny.

We used VOSviewer to visually analyze the distribution of countries/regions, journals, and keywords. After importing the data, we selected “Co-authorship” with “Countries” as the unit of analysis to examine international collaboration, “Bibliographic coupling” with “Sources” to analyze journal distribution, “Co-occurrence” with “Author keywords” to identify research hotspots, and “Co-citation” with “Cited sources” to explore citation relationships among journals. All analyses in VOSviewer were performed using the full counting method and the association strength normalization method. The minimum thresholds were set as follows: 5 documents for country-level analysis, 5 occurrences for keyword co-occurrence, and 20 citations for journal co-citation. VOSviewer automatically grouped items into clusters, each represented by a different color, with each cluster reflecting a distinct research theme or topic. These analyses aimed to explore international collaboration, identify emerging research trends, and uncover structural and thematic patterns in the literature. Compared to CiteSpace—which focuses more on temporal patterns and burst detection—VOSviewer is particularly effective in revealing structural relationships and network density. During the visualization process, some data cleaning and standardization were applied. Specifically, records from Taiwan and Hong Kong were merged under China, and data from England, Scotland, Wales, and Northern Ireland were combined under the United Kingdom to ensure consistency in country classification.

In addition, we used CiteSpace to visualize institutional collaboration networks, journal dual-map overlays, co-cited references, citation bursts, keyword clustering, and keyword bursts. Before importing the data, we created a folder containing four subfolders named “data,” “input,” “output,” and “project.” The records exported from Web of Science were placed in the “input” folder and deduplicated before analysis. The CiteSpace parameters were configured as follows: (1) time span: January 2014 to December 2023; (2) years per slice: 1; (3) term sources: Title, Abstract, Author Keywords, and Keywords Plus; (4) node types: Institution, Keyword, Reference; (5) link strength: Cosine; scope: Within slices; (6) selection criteria: g-index (k = 25), Top N per slice = 50, Top N% = 10%; (7) pruning method: Pathfinder; (8) visualization: Cluster View—Static with “Show Merged Network” enabled. Institutional analysis was used to identify key research organizations and collaboration patterns; the dual-map overlay revealed interdisciplinary citation relationships; co-citation analysis helped locate core literature in the field; citation burst analysis identified studies that gained increased scholarly attention in specific periods; keyword clustering reflected thematic structures; and keyword burst analysis captured emerging research hotspots. Compared to VOSviewer, which focuses more on static structural visualization, CiteSpace is more effective in detecting temporal trends, citation bursts, and critical turning points in the evolution of scientific knowledge.

Co-citation frequency refers to the number of times two documents are cited together by a third-party publication. In co-citation analysis, a higher frequency indicates a closer scholarly relationship between the two items and suggests that they may form part of the intellectual foundation of a research field. Therefore, co-citation frequency is commonly used to rank authors, journals, or references in order to identify influential contributions within the academic community. ²⁵

Results

Published papers and their development trends

We retrieved a total of 2169 publications related to the pathological mechanisms of AD using mouse models, including 2056 articles and 113 reviews. Articles represented 95% of the total and reviews accounting for 5%. To evaluate publication trends from 2014 to 2023, we generated the bar chart presented in Figure 2. The overall number of publications during this period exhibited a pattern of decline followed by a subsequent increase. The peak occurred in 2015, with 290 publications, representing approximately 13% of the total. From 2014 to 2018, the number of publications steadily declined, reaching a low of 133 in 2018, or about 6.1% of the total. However, from 2018 to 2023, publication numbers fluctuated but showed an overall upward trend. Indicating a resurgence of interest in this area among researchers in recent years.

Figure 2.

The bar chart illustrates the publication trend in this field from 2014 to 2023.

Countries/regions

A total of 69 countries and regions have contributed in research on the pathological mechanisms of AD using mouse models. Figure 3A illustrates the global collaboration network for AD research using mouse models, highlighting the USA as the central hub connecting major contributors such as Europe and China. The thickness of the links represents the strength of collaborative relationships between countries or regions, with the thickest line observed between the USA and China, indicating the highest frequency of collaboration. The color intensity of each country reflects its publication output, where darker shades correspond to higher numbers of publications. Table 1 lists the top 10 countries/regions ranked by publication volume and citation frequency. The USA leads with 943 publications, followed by China (426), Germany (194), and the United Kingdom (148). In terms of citation frequency, the USA far exceeds other countries, with 49,310 citations. Additionally, China (12,604), Germany (8500), the United Kingdom (8349), and Japan (5678) also exhibit high citation frequencies, while other countries have fewer than 5000 citations.

Figure 3.

Analysis of countries/regions in the field. (A) A global collaboration map generated using the Bibliometrix R package. Each line represents collaborative relationships between countries/regions based on data from the Web of Science Core Collection (2014–2023). The thickness of the lines indicates the strength of collaboration, while the color intensity reflects the number of publications from each country. (B) A global collaboration network visualization created with VOSviewer. The size of each node represents the number of publications from that country/region, and the links between nodes indicate collaborative relationships. Different colors distinguish clusters of countries with close collaborative ties.

Table 1.

Top 10 countries/regions ranked by publication volume and citation frequency.

Rank	Country	Documents	Rank	Country	Citations
1	USA	943	1	USA	49,310
2	China	443	2	China	13,076
3	England	174	3	England	9177
4	Germany	194	4	Germany	8500
5	Japan	144	5	Japan	5678
6	Canada	142	6	Canada	4974
7	Spain	130	7	Spain	4623
8	France	87	8	France	3741
9	South Korea	83	9	Belgium	3343
10	Australia	78	10	South Korea	3083

Figure 3B depicts the collaboration network among countries involved in AD pathological mechanism research using mouse models, displaying countries with five or more publications. In VOSViewer, countries and regions are primarily divided into 4 blocks, represented by different colors based on the closeness of cooperation. The green block includes countries like the USA, China, Japan, Singapore, South Korea, and Australia. The red block includes the Czech Republic, Finland, France, Germany, Hungary, Poland, and Switzerland. The blue blocks mainly include countries such as Austria, Egypt, England, Ireland, Italy, and Saudi Arabia. The yellow blocks mainly include countries such as Argentina, Brazil, Canada, Chile, Mexico, and Spain. Each circle represents a different country/region, with the size of the circle indicating the number of publications from that country/region. The proximity of the circles reflects the strength of collaboration, with closer circles indicating stronger ties. The thickness of the connecting lines signifies the closeness of collaboration, while the color of the circles indicates clusters of collaborating countries. These results highlight the USA, China, Germany, and the UK as the core of a global collaboration network, linking multiple countries and forming a collaboration framework centered on the USA, with regional clusters in Europe and Asia.

Institutions

We employed CiteSpace software to construct a network map of institutional collaborations in this field (as shown in Figure 4). The network comprises 361 nodes and 1744 links, where the size of each node corresponds to the number of publications by the respective institution, and the connections between nodes represent collaborative relationships among institutions. Purple rings around nodes indicate high betweenness centrality, with deeper purple color reflecting greater centrality.

Figure 4.

Analysis of institutions in the field using citeSpace. Each node represents a research institution, with its size proportional to the number of publications produced by that institution. The connections between nodes indicate collaborative relationships among institutions. Purple rings around nodes indicate high betweenness centrality, with deeper purple color reflecting greater centrality.

Table 2 presents the top 10 institutions ranked by publication volume and centrality. The University of California System published the most articles (116), followed by Harvard University (81), the University of Texas System (69), the State University System of Florida (65), the Helmholtz Association (64), and the Veterans Health Administration (62). The publication counts of the remaining institutions are all below 60.

Table 2.

Top 10 institutions ranked by publication volume and centrality.

Rank	Institution	Publications	Rank	Institution	Centrality
1	University of California System	116	1	Harvard University	0.17
2	Harvard University	81	2	University of California System	0.14
3	University of Texas System	69	3	Chinese Academy of Sciences	0.14
4	State University System of Florida	65	4	University of Texas System	0.13
5	Helmholtz Association	64	5	State University System of Florida	0.12
6	Veterans Health Administration (VHA)	62	6	Veterans Health Administration (VHA)	0.09
7	US Department of Veterans Affairs	62	7	US Department of Veterans Affairs	0.09
8	German Center for Neurodegenerative Diseases (DZNE)	54	8	Consejo Superior de Investigaciones Cientificas (CSIC)	0.09
9	Institut National de la Sante et de la Recherche Medicale (Inserm)	53	9	Washington University (WUSTL)	0.08
10	Massachusetts General Hospital	50	10	Massachusetts General Hospital	0.07

The fourth column of Table 2 lists the top 10 institutions by centrality. Harvard University has the highest centrality score (0.17), followed by the University of California System (0.14), the Chinese Academy of Sciences (0.14), the University of Texas System (0.13), and the State University System of Florida (0.12). Notably, these institutions are marked with purple rings on their nodes, indicating centrality scores above 0.1.

It is noteworthy that the University of California System, Harvard University, the University of Texas System, and the State University System of Florida rank among the top five in both publication volume and centrality, underscoring their pivotal role as intermediaries in the international collaboration network.

Betweenness centrality specifically measures a node's ability (such as an author, institution, or country/region) to serve as a bridge or intermediary between other nodes within the network. In CiteSpace analysis, a centrality score greater than 0.1 is typically considered indicative of a key or highly influential node in the network.

Journals

A total of 373 academic journals have published articles related to this field. Table 3 presents the top 10 journals ranked by publication volume and co-citation frequency. The journal with the highest number of publications is Journal of Alzheimer's Disease (190, Q2, IF 3.4), followed by Neurobiology of Aging (77, Q2, IF 3.7) and International Journal of Molecular Sciences (69, Q1, IF 4.9).

Table 3.

Top 10 journals ranked by publication volume, co-citation frequency, impact factor (IF, JCR 2021), and JCR quartile ranking.

Rank	Journals	Publications	Rank	Journals	Co-citations
1	Journal of Alzheimer's Disease 3.4 (Q2)	190	1	Journal of Neuroscience 4.4 (Q1)	7215
2	Neurobiology Of Aging 3.7 (Q2)	77	2	PNAS 9.4 (Q1)	4122
3	International Journal of Molecular Science 4.9 (Q1)	69	3	Neurobiology of Aging 3.7 (Q2)	3906
4	Neurobiology of Disease 5.1 (Q1)	61	4	Neuron 14.7 (Q1)	3799
5	Frontiers In Aging Neuroscience 4.1 (Q2)	57	5	Journal of Biological Chemistry 4.0 (Q2)	3725
6	Acta Neuropathologica Communications 6.2 (Q1)	55	6	Journal of Alzheimer's Disease 3.4 (Q2)	3616
7	Scientific Reports 3.8 (Q1)	52	7	Science 44.7 (Q1)	2842
8	Molecular Neurodegeneration 14.9 (Q1)	48	8	Nature 50.5(Q1)	2716
9	Journal of Neuroinflammation 9.3 (Q1)	48	9	PLoS One 2.9 (Q3)	2591
10	Journal of Neuroscience 4.4 (Q1)	45	10	Journal of Neurochemistry 4.2 (Q2)	2242

The journal with the highest number of co-citations is Journal of Neuroscience (7,251, Q1, IF 4.4), followed by Proceedings of the National Academy of Sciences of the United States of America (PNAS) (4,122, Q1, IF 9.4) and Neurobiology of Aging (3,906, Q2, IF 3.7). Notably, Neurobiology of Aging ranks in the top three for both publication volume and co-citation frequency, underscoring its significant influence in this field.

Figure 5 visualizes the relationships among journals publishing research on the pathological mechanisms of AD using mouse models. In VOSviewer, journals are categorized into four distinct, color-coded clusters based on their thematic similarity. The red cluster includes journals such as Journal of Alzheimer's Disease, International Journal of Molecular Sciences, and Frontiers in Aging Neuroscience, which focus on AD and aging mechanisms, such as those explored in Neurobiology of Aging and Frontiers in Aging Neuroscience. The green cluster comprises Molecular Neurodegeneration, Journal of Neuroscience, and Journal of Neuroinflammation, with a primary focus on the neuropathological mechanisms of AD, as seen in Journal of Neuroscience and Journal of Neuroinflammation. The blue cluster features Acta Neuropathologica Communications, Neurobiology of Disease, and Alzheimer's Research & Therapy, emphasizing therapeutic strategies for AD, including research published in Alzheimer's Research & Therapy and Theranostics. The yellow cluster contains fewer journals with more dispersed research areas, lacking a clear focal direction.

Figure 5.

Analysis of journals in the field. (A) A journal collaboration network visualization created with VOSviewer. The size of each node represents the number of publications for each journal, and the links between nodes indicate collaboration relationships. Different colors represent clusters of journals with frequent collaborations. (B) A journal co-citation network visualization created with VOSviewer. Node size reflects the total co-citation frequency of each journal, and the links indicate how often journals are cited together. Different colors indicate clusters of journals that are commonly co-cited. (C) A journal dual-map overlay generated with CiteSpace. The map displays citing journals on the left and cited journals on the right. Each ellipse represents a journal, where vertical length indicates the number of publications and horizontal length indicates the number of contributing authors. Colored curved lines represent citation paths, illustrating interdisciplinary citation relationships across research domains.

Based on co-citation frequency analysis, journals in this field can also be grouped into four clusters with similar research directions (Figure 5B). The red cluster includes journals such as Journal of Alzheimer's Disease and Journal of Neurochemistry. The green cluster includes Journal of Neuroscience, Neuron, and Neurobiology of Aging. The yellow cluster features Acta Neuropathologica, NeuroImage, and Brain, primarily focusing on neuroscience and neuropathology studies. The blue cluster includes leading journals like Journal of Biological Chemistry, PNAS, Nature, Science, and Cell, which emphasize biochemical and molecular biological research on AD.

The journal dual-map overlay illustrates the positioning of research on this topic relative to major research disciplines, as shown in Figure 5C. Each point on the map represents a journal, with the map divided into two parts: the citing map on the left and the cited map on the right. Curved lines indicate citation links, fully tracing the citation flow. Ellipses represent the number of publications for each journal, displaying the ratio between authors and publications. The length of the ellipse corresponds to the number of authors, while its width reflects the number of publications (the longer the vertical axis, the more papers the journal has published; the longer the horizontal axis, the more authors it has). The curves between the left and right parts of the map are citation links, providing insight into the interdisciplinary relationships of the field. Publications in the fields of molecular biology, immunology, medicine, and clinical sciences are significantly influenced by research from a broad range of disciplines, including molecular biology, genetics, technology, metallurgy, veterinary sciences, animal studies, parasitology, health sciences, nursing, medicine, psychology, education, social sciences, chemistry, materials science, physics, environmental science, toxicology, and nutrition. These citation pathways indicate that AD research using mouse models is inherently interdisciplinary, drawing heavily from fields such as molecular biology and genetics, and influencing clinical medicine and health sciences.

Authors and cited authors

This study analyzed 13,632 authors contributing to 2169 articles. Figure 6A and Figure 6B present the top 20 authors based on publication count and citation frequency. The most prolific author is Saito T (34 publications), followed by Saido TC (31) and Holtzman DM (28). Holtzman DM is also the most cited author (4434 citations), followed by Hyman BT (2204) and Lee VMY (2098), underscoring Holtzman DM's significant influence in the field. This highlights Holtzman DM's significant influence in the field. Figure 6C illustrates the publication and citation trends of the top 20 authors from 2014 to 2023. The circle size represents publication volume, with larger circles indicating higher output, while the color intensity reflects citation frequency, with darker colors signifying higher citations. Holtzman DM consistently produced high-quality publications annually, whereas Saito T and Saido TC, despite their high publication counts, need to improve their citation frequencies.

Figure 6.

Analysis of leading authors in the field using the bibliometrix R package. (A) Top 20 Authors by Publication Count. (B) Top 20 Authors by Citation Frequency. (C) Trends in Annual Publications and Citation Frequencies (2014–2023) for the Top 20 Authors. Circle size reflects the number of publications, while color intensity represents the citation impact of each author.

Co-cited author analysis identifies instances where the works of two authors are simultaneously cited by a third author. Higher co-citation frequencies indicate closer academic interests and research relevance (35). Table 4 lists the top 10 co-cited authors by frequency. The most co-cited author is Selkoe DJ (532), followed by Oddo S (461) and Braak H (439), underscoring Selkoe DJ's leadership in the field.

Table 4.

Top 10 authors by co-citation frequency.

Rank	Author	Citations
1	Selkoe, DJ	532
2	Oddo, S	461
3	Braak, H	439
4	Jankowsky,JL	399
5	Heneka, MT	346
6	Hardy, J	316
7	Oakley, H	306
8	Palop, JJ	239
9	Goedert,M	203
10	Jack, CR	180

Cited references

The citation network generated using CiteSpace includes 781 nodes and 3862 links. Each node represents a citation, and its size is positively correlated with citation frequency—the more cited a publication is, the larger its node. Links between nodes indicate that these references have been co-cited in the same paper, with the thickness of the lines reflecting the strength of their co-citation relationship. Node colors represent citation periods: purple indicates earlier citations, while red signifies recent citations.

As shown in Figure 7A, the nodes for Hardy J (2002), Oddo S (2003), and Oakley H (2006) are notably large, reflecting their significant contributions to AD pathological mechanism research using mouse models. Among these, the blue areas of Hardy J and Oddo S nodes are larger, indicating frequent citations around 2015. In contrast, Oakley H has a smaller blue area and a larger red area, signifying a surge in citations in recent years. Additionally, Keren-Shaul H (2017) is another prominent node with a large red area, highlighting its growing importance in recent years.

Figure 7.

Analysis of references in the field using citeSpace. (A) Bibliometric analysis of co-cited references in this field. Each node represents a cited reference, with node size indicating the total number of citations. The color intensity reflects the citation impact. (B) Burst analysis of references in the field. Each horizontal bar represents a reference, with the red segment indicating the period during which the reference experienced a surge in citations.

Table 5 lists the most cited references in the field. In Table 5, the most frequently cited reference is “The amyloid hypothesis of Alzheimer's disease: progress and problems on the road to therapeutics”, published in Science in 2002, with 224 citations. ²⁶ Notably, we evaluated the centrality of the cited references using CiteSpace, and found that the centrality of all cited references was less than 0.1. References with higher centrality are often capable of connecting different research groups, significantly influencing interdisciplinary development and knowledge dissemination in the field.

Table 5.

Top 10 most-cited references in terms of total citations.

Rank	Frequency	Title	Journal	JCR	IF
1	224	The amyloid hypothesis of Alzheimer's disease: progress and problems on the road to therapeutics. 26	Science	Q1	56.9
2	179	Mutant presenilins specifically elevate the levels of the 42 residue beta-amyloid peptide in vivo. 27	Human Molecular Genetics	Q2	5.1
3	168	Neuropathological stageing of Alzheimer-related changes. 28	Acta Neuropathologica	Q1	12.7
4	144	Correlative memory deficits, Abeta elevation, and amyloid plaques in transgenic mice. 29	Science	Q1	56.9
5	142	Synapse loss and microglial activation precede tangles in a P301S tauopathy mouse model. 30	Neuron	Q1	16.2
6	129	Neuroinflammation in Alzheimer's disease. 31	Lancet Neurology	Q1	46.2
7	125	Tau suppression in a neurodegenerative mouse model improves memory function. 32	Science	Q1	56.9
8	122	Alzheimer's disease: the amyloid cascade hypothesis. 33	Science	Q1	56.9
9	121	Alzheimer's disease. 1	New England Journal of Medicine	Q1	158.5
10	120	The amyloid hypothesis of Alzheimer's disease at 25 years. 34	EMBO Molecular Medicine	Q1	12.3

In CiteSpace, the term “burst” refers to the phenomenon in which the citation frequency of a paper sharply increases within a specific time period. Burst references represent research outcomes that have gained widespread attention in the field during a particular phase, often indicating significant academic recognition and influence over a short period. ³⁵

Figure 7B illustrates the 25 most influential burst citations over the past decade, as generated by CiteSpace. The blue bar charts represent the time span of the cited references from 2014 to 2023, while the red bar charts indicate the burst periods, during which the citation count of these references significantly increased. Among the recent burst references, the most recent paper is “Inflammation as a central mechanism in Alzheimer's disease,” published in Alzheimer's & Dementia (New York) in 2018, with a burst period from 2021 to 2023. ³⁶

In burst reference analysis, “intensity” refers to the magnitude or extent of the increase in citations of a specific paper or keyword within a given time frame. ³⁷ A higher intensity value signifies a greater impact or a more significant increase in citations during the burst period.

As shown in Figure 7B, the highest intensity is attributed to the article by Hadas Keren-Shaul, titled “A Unique Microglia Type Associated with Restricting Development of Alzheimer's Disease,” published in Cell in 2017. This article experienced a burst period from 2019 to 2023. The activity pathways and molecular mechanisms of microglia in different stages of AD remain a topic of debate. This study, which employed single-cell RNA sequencing technology, comprehensively mapped immune cell populations in the brains of wild-type and AD transgenic (Tg-AD) mice, identifying a novel microglial type (DAM) linked to neurodegenerative diseases. ³⁸ The paper provides valuable insights into the pathological mechanisms of AD using mouse models and proposes potential research directions for future AD treatment strategies. Notably, this paper ranks 14th in citation frequency among the most-cited references (2017), and Keren-Shaul H (2017) appears as a prominent node in Figure 7A, with a large red area, indicating the significant academic influence and widespread recognition of this work in the field of AD pathology mechanisms based on mouse models in recent years.

Keywords

A word cloud was generated using the R package “ Bibliometrix”, as shown in Figure 8A. The keywords are categorized into three main areas: (1) Keywords related to pathology and biology, such as “pathology,” “A-beta,” “tau,” “neuroinflammation,” “neurodegeneration,” “inflammation,” “oxidative stress,” “microglia,” “protein,” “activation,” and “expression. “ (2) Keywords related to animal models, such as “mouse model,” “transgenic mice,” “mice,” and “in-vivo.” (3) Keywords related to research areas and clinical features, such as “brain,” “hippocampus,” “memory,” and “dementia.”

Figure 8.

Analysis of keywords in the field. (A) Keyword word cloud generated using the Bibliometrix R package. (B) Keyword co-occurrence network visualization generated with VOSviewer. Each node represents a keyword, and the size of the node reflects its frequency of occurrence in the literature. The links between nodes indicate co-occurrence relationships, with thicker lines representing higher co-occurrence frequency. Different colors represent thematic clusters automatically identified by the algorithm. (C) Cluster analysis of keywords generated with CiteSpace. Colors represent different clusters, displaying the nine largest clusters.

Figure 8B displays the co-occurrence network of these keywords, with closely related keywords clustered into categories, represented by five large colored blocks. The red cluster relates to amyloid pathology, including keywords such as “amyloid-beta,” “amyloid plaques,” “BACE1,” and “gamma-secretase,” reflecting the central role of Aβ in AD. The green cluster focuses on cognitive function and synaptic plasticity, with terms like “cognition,” “synaptic plasticity,” “learning and memory,” and “BDNF.” The blue cluster is associated with tau pathology, including “tau,” “tau phosphorylation,” “tauopathy,” and “tau immunotherapy.” The yellow cluster highlights neuroinflammation and immune activation, with terms such as “neuroinflammation,” “microglia,” and “TREM2.” The purple cluster pertains to metabolic dysfunction and oxidative stress, with keywords like “insulin resistance,” “diabetes,” and “mitochondrial dysfunction.” Notably, Aβ protein is central across all five research blocks, reflecting its key role in AD pathology.

Using CiteSpace, we generated nine clusters. In network analysis and clustering, the mean silhouette score (S) and the modularity score (Q) are key indicators used to assess the quality of clusters and communities. Generally, S ≥ 0.5 indicates that clustering is meaningful, and Q ≥ 0.3 suggests that community detection is effective. As shown in Figure 8C, Q = 0.3184 and S = 0.6441, both of which meet the criteria, confirming the robustness of our clustering network results. In Figure 8C, there are 9 labels representing 9 clusters. Each cluster label corresponds to the most popular term in the co-occurrence network, and the clusters are numbered from #0 to #8. The cluster numbering is inversely related to the number of documents included in the cluster—the larger the number, the fewer documents the cluster contains. The 9 clusters are as follows: #0 Alzheimer's disease, #1 Insulin resistance, #2 Amyloid beta, #3 Neurodegenerative diseases, #4 Therapeutic ultrasound, #5 Human APP Swedish, #6 Prepulse inhibition, #7 Drug repurposing, and #8 Oxidative stress.

The annual heatmap provides a chronological overview of keyword usage trends over the past decade, highlighting the evolution of research focus within the field. As shown in Figure 9A, the Value reflects the relative importance or attention of a specific keyword in a given year (the higher the citation count, the brighter the color). Keywords such as amyloid-beta and Alzheimer's disease remained research hotspots throughout the entire period, consistently displaying high brightness. In contrast, keywords like astrocytes and synaptic plasticity have shown increasing brightness in recent years, indicating that these areas may represent emerging research focuses. Meanwhile, keywords such as oxidative stress and neuroinflammation have gradually diminished in brightness in certain years, potentially indicating a decline in research interest in these fields. Figure 9B presents the keyword burst analysis. The keyword “Alzheimer's disease” exhibits the highest burst strength, reaching 17.62, underscoring its central role in this study. Additionally, other keywords such as Pathogenesis, Sex differences, National institute, Gut microbiota, NLRP3 inflammasome, System, Metabolism, Plasticity, Alzheimer's disease, and Damage are still in the burst phase and continue to have high attention and impact in 2023. The presence of these burst keywords suggests that areas such as AD pathogenesis, sex differences, the microbiome, metabolism, and other related fields are emerging as new research hotspots, attracting sustained academic interest and further exploration.

Figure 9.

Analysis of keyword heatmap in the field. (A) Annual heatmap of keywords from 2014 to 2023. Each row represents a keyword, and the color intensity across years indicates its relative importance or level of attention in that year; brighter colors represent higher attention. (B) Burst detection analysis of keywords. Each horizontal bar represents a keyword, with the red segment indicating the period during which the keyword experienced a surge in citations.

Quantitative characteristics of the bibliometric networks

To provide a more comprehensive understanding of the structural properties of the bibliometric networks, we calculated several standard network metrics, including the number of nodes, links, clusters, density, total link strength, modularity (Q), silhouette (S) and average degree. A summary of these quantitative indicators is presented in Table 6.

Table 6.

Summary of quantitative metrics for the bibliometric networks constructed using VOSviewer and CiteSpace.

Figure	Network Type	Software	I	L	D	A	C	T	Q	S
Figure 3B	Country Collaboration	VOSviewer	38	268	0.38	14.11	6	1254	—	—
Figure 4	Institution Collaboration	CiteSpace	361	1744	0.03	9.66	—	—	—	—
Figure 5A	Journal Collaboration	VOSviewer	97	4633	1.00	95.53	4	436,857	—	—
Figure 5B	Journal Co-citation	VOSviewer	630	143,271	1.03	10.66	5	4,648,241	—	—
Figure 7A	Reference Co-citation	CiteSpace	781	3862	0.01	9.89	—	—	—	—
Figure 8B	Keyword Co-occurrence	VOSviewer	284	4130	2.03	11.66	14	9199	—	—
Figure 8C	Keyword Clustering	CiteSpace	587	4404	0.03	15.01	9	—	0.3184	0.6441

Notes: Figure refers to the corresponding figure number in the paper. Network Type indicates the type of network analyzed, such as country collaboration or journal co-citation. Software specifies the tool used for analysis, either VOSviewer or CiteSpace. I (Items) represents the number of nodes in the network (e.g., countries, institutions, journals, or keywords), while L (Links) denotes the number of connections between these nodes. D (Density) measures how interconnected the network is, with values ranging from 0 to 1. A (Average Degree) indicates the average number of connections per node. C (Clusters) shows the number of distinct communities or groups detected within the network. T (Total Link Strength), available in VOSviewer, reflects the overall intensity of collaboration or co-occurrence. Q (Modularity Q), provided by CiteSpace, assesses the strength of the division of the network into modules, with higher values indicating clearer clustering structures. S (Silhouette S), also available in CiteSpace, indicates the quality and consistency of the clustering, with values closer to 1 suggesting better-defined clusters.

Discussion

This study is the first bibliometric and visual analysis of AD pathophysiological mechanisms based on mouse models. We retrieved 2169 publications published between 2014 and 2023 that are closely related to this field. By analyzing various aspects of the literature, including publication volume, countries/regions, journals, institutions, authors, co-cited references, and keywords, we found that AD is globally prevalent and significantly impacts the quality of life. This field has attracted increasing attention, drawing more and more scholars to conduct research.

Researchers from 69 countries are engaged in AD research. Among the top 10 contributing countries, the United States leads not only in the number of publications but also in total citations, indicating its prominent position in terms of academic influence and publication quality. The USA has created a globally dominant collaboration centered around it.

Harvard University stands out as the central institution in the international collaboration network of this field, with the highest centrality. The University of California System, Harvard University, the University of Texas System, and the State University System of Florida are ranked among the top five institutions in both publication volume and centrality, demonstrating their significant influence in the field.

The journals in this field predominantly focus on areas such as molecular biology, immunology, medicine, and clinical sciences. These disciplines are further influenced by other fields, including genetics, technology, metallurgy, veterinary sciences, animal studies, parasitology, health sciences, nursing, psychology, education, social sciences, chemistry, materials science, physics, environmental science, toxicology, and nutrition. The Journal of Alzheimer's Disease has the highest publication volume, while the Journal of Neuroscience has the highest total citation frequency. Notably, Neurobiology of Aging ranks in the top three for both publication volume and co-citation frequency, underscoring its strong impact in the field.

Journals can be categorized based on their focus areas, with clusters primarily dedicated to AD and aging mechanisms, neuropathological mechanisms of AD, and therapeutic strategies for AD. Co-citation analysis of journals reveals that the primary areas of focus are neuroscience, neuropathology, and biochemical and molecular biology.

Holtzman DM is a key figure in this field, producing high-quality publications almost every year. While Saito T and Saido TC rank highly in publication volume, their citation frequency still has room for improvement. Selkoe, DJ is the most co-cited author.

By analyzing the co-occurrence of keywords in the literature, the main themes and research hotspots in a particular field can be identified. ³⁹ From the keyword word cloud and co-occurrence analysis, we can see that the keywords mainly fall into five categories: research related to synaptic and neuronal function in AD, AD-related blood-brain barrier and cerebrovascular pathology, tau protein and neurofibrillary tangles, AD and metabolic disorders, and diagnostics and biomarkers.

From the keyword clustering and burst analysis, we observed that research hotspots have gradually expanded from in-depth studies on classic pathology (such as Aβ and tau protein) to more systemic pathological mechanisms. “Aβ” and “tau” have remained central across multiple clusters, reflecting their fundamental position in AD research. Emerging pathological hotspots include synapse loss, gut microbiota, and NLRP3 inflammasome. We will now discuss these three emerging hotspots in detail.

In the keyword burst analysis of AD pathological mechanisms research based on mouse models, we identified “synapse loss” as the top-ranked keyword with a strength of 5.37. This indicates that synapse loss occupies a central position in AD research. Synapses are the structures that enable communication between neurons or between neurons and effector cells. They are critical sites for information transmission in the central nervous system, allowing electrical or chemical signals to pass from one cell to another. ⁴⁰ The basic structure of a synapse includes the presynaptic membrane, synaptic cleft, postsynaptic membrane, and synaptic vesicles. Beyond signal transmission, synapses are involved in learning and memory processes, adjusting the strength of signal transmission through synaptic plasticity, such as long-term potentiation (LTP) and long-term depression (LTD) ⁴¹ . This dynamic nature of synapses underpins the brain's adaptability and learning capacity. AD is a neurodegenerative disorder characterized by brain deposition of amyloid plaques and tau neurofibrillary tangles, along with progressive cognitive decline.^42–46 During AD progression, one notable change in synapses is synapse loss. Studies on cortical neurons cultured from AD transgenic mice (APP/PS1) have shown a significant reduction in spine density, indicating a decreased potential for synapse formation or connectivity in neurons of AD mice. ⁴⁷ Synaptic dysfunction is a direct cause of cognitive decline, revealing the pathological process of impaired communication between neurons. Growing evidence has highlighted the close relationship between synapse loss and the onset and progression of AD. Below, we summarize the pathological mechanisms underlying synapse loss in AD.

Intracellular Amyloid Beta (Aβ) Oligomers Induce Synapse Loss and Degeneration, Leading to AD Onset: Research utilizing embryos (E18) of wild-type (MAPT+/+) and tau-knockout (MAPT−/−) mice has cultured primary neurons and analyzed synapses. The results demonstrated that intracellular Aβ oligomers reduce the number of mature dendritic spines, and this effect is tau-independent. Aβ oligomers impair intracellular transport mechanisms essential for synapse maintenance, including brain-derived neurotrophic factor, mitochondria, and recycling endosomes, thus compromising synaptic function. The study also observed that intracellular Aβ oligomers are deposited earlier than extracellular Aβ in AD model mice, suggesting that intracellular Aβ oligomers may contribute to the early synaptic pathology of AD. ⁴⁸ Synaptic strength can be regulated by LTP and LTD. Aβ oligomers disrupt synaptic function by altering the balance of LTP and LTD, promoting synapse loss.^49,50 Aina Ng et al. ⁵¹ treated mouse hippocampal slices with Aβ_1–42 oligomers and recorded excitatory synaptic transmission using electrophysiological techniques. They found that the mGluR-dependent LTD, mediated by metabotropic glutamate receptors (mGluRs), was disrupted in AD mouse models. Exogenous application of Aβ_1–42 oligomers to mouse hippocampal slices enhanced mGlu5R-dependent LTD. Moreover, enhanced synaptic weakening was mediated through N-methyl-D-aspartate receptors (NMDARs) and complement C5aR1 signaling. These interactions among mGlu5R, NMDAR, and the complement system in Aβ_1–42 oligomer-induced synaptic weakening may represent early triggers for AD-associated synapse loss and degeneration.

Tau Protein-Induced Synapse Degeneration and Loss, Leading to AD Onset: Tauopathy mouse models have been employed to identify the toxic effects of tau on synapses, dendritic spines, and neuronal function. Increasing evidence suggests that tau is involved in synapse loss in AD. ⁵² By regulating the expression of mutated tau protein (TauRD) in transgenic mice, studies evaluated the effects of TauRD expression on neuronal motor function, behavior, learning, memory, and synaptic plasticity. Results showed severe memory impairment in aggregation-prone TauRD mice, presenting AD-typical pathological changes, including aggregation, mislocalization, hyperphosphorylation, synapse loss, and neuron loss. Continuous presence of aggregation-prone TauRD was identified as the primary toxic form affecting memory and LTP rather than the aggregates themselves. ⁵³ Interestingly, the localization of tau within synapses suggests that tau may play a role in normal synaptic function in both healthy and AD mouse brains. However, normal synaptic functions may be disrupted in disease states. Tau can directly interact with postsynaptic signaling complexes, regulate glutamate receptor content in dendritic spines, and influence synaptic mitochondrial targeting and function, ultimately contributing to AD onset. ⁵⁴

Aberrant Microglial Phagocytosis Causes Synapse Loss in AD: Mutations in the microglial protein TREM2 (triggering receptor expressed on myeloid cells 2) have been linked to an increased risk of AD, with the R47H/+ substitution being particularly notable. Jay Penney et al. ⁵⁵ utilized gene editing and stem cell models to investigate the effects of the TREM2 R47H/+ mutation on human iPSC-derived microglia. They found that mouse brains transplanted with TREM2 R47H/+ microglia exhibited reduced synaptic density. The TREM2 R47H/+ mutation was associated with several detrimental effects on microglial gene expression and function, which may underlie its connection to AD. Additionally, studies have shown elevated AIM2 (absent in melanoma 2) expression in the microglia of AD mice. Conditional knockout of AIM2 in microglia rescued cognitive and synaptic dysfunction in AD mice. ⁵⁶

In summary, synapse loss is a prominent topic in AD pathological mechanism research based on mouse models and a core aspect of AD pathology, involving interactions among pathological proteins, synapses, and glial cells. Understanding these mechanisms is crucial for developing new therapeutic strategies and biomarkers for AD.

The second-ranking keyword, “gut microbiota,” with a burst strength of 4.49, highlights its critical role in exacerbating AD progression through various mechanisms. Gut microbiota dysbiosis increases lipopolysaccharide (LPS) levels, which activate immune cells in the central nervous system and induce microglial hyperactivation, thereby intensifying neuroinflammation and promoting Aβ aggregation.^57,58 Additionally, gut microbial metabolites such as short-chain fatty acids (SCFAs) play an essential role in regulating neuroinflammation and enhancing Aβ clearance. Studies indicate that reduced SCFA levels suppress neurogenesis, leading to hippocampus-dependent behavioral impairments.^59,60 Research also highlights the importance of interactions between gut microbiota and host genotypes in the pathological mechanisms of AD. For example, the gut microbiota of ApoE4 genotype mice tends to produce pro-inflammatory metabolites (e.g., LPS), worsening neuroinflammation and cognitive dysfunction, while the ApoE3 genotype enhances SCFA production, reducing Aβ aggregation and improving cognitive performance. ⁶⁰ Moreover, gut-brain axis studies have unveiled various potential strategies for modulating gut microbiota. Fecal microbiota transplantation (FMT) has proven highly effective in alleviating AD symptoms. For instance, FMT experiments have demonstrated that transplanting healthy donor gut microbiota into AD mouse models significantly reduced brain Aβ deposition and improved learning and memory abilities. ⁶¹ Additionally, interventions such as intermittent fasting and natural compounds (e.g., camellia oil) have been shown to mitigate AD pathology by modulating gut microbiota and their metabolites. Intermittent fasting notably increased gut microbiota diversity and improved Aβ clearance and cognitive function through polyunsaturated fatty acid (PUFA) metabolism. ⁶²

Regulating the composition and function of gut microbiota presents promising new avenues for AD prevention and treatment.

Finally, the third-ranking keyword, “inflammasome,” highlights the critical role of neuroinflammation in the pathogenesis of AD. ⁶³ Among inflammasomes, the activation of the NLRP3 inflammasome is considered a key mechanism exacerbating Aβ and tau pathology. ⁶⁴ As an intracellular multiprotein complex, its activation not only promotes Aβ deposition but also aggravates tau hyperphosphorylation and neuronal loss, serving as a critical link between Aβ plaques and tau pathology. ⁶⁵ Studies have shown that Aβ can induce NLRP3 inflammasome activation by upregulating NLRP3 and IL-1β expression, thereby enhancing inflammatory responses. ⁶⁶ The activation of the NLRP3 inflammasome plays a significant role in neuroinflammation and profoundly impacts the functional state of microglia. Under normal conditions, microglia in their M2 anti-inflammatory phenotype participate in Aβ clearance and neuroprotection. However, excessive activation of NLRP3 shifts microglia toward the M1 pro-inflammatory phenotype, leading to the release of large amounts of pro-inflammatory factors such as IL-1β and TNF-α, while markedly impairing their ability to clear Aβ. ⁶⁷ This indicates that NLRP3 inflammasome activation is central to AD pathogenesis, making it an emerging therapeutic target. Targeting the degradation of the NLRP3 inflammasome through the autophagy-lysosome pathway has been shown to significantly improve AD pathological features. This approach maintains the phagocytic function of microglia, reduces Aβ plaque deposition and neuroinflammation, and enhances cognitive function. In NLRP3 knockout mice, the size and number of Aβ plaques in the hippocampus and frontal cortex were significantly reduced, and the phagocytic function of microglia was restored. ⁶⁸ Targeted inhibition of NLRP3 inflammasome activation reduces microglial (MG) M1 polarization and neuroinflammation, thereby improving cognitive function in AD mice. ⁶⁹ Furthermore, mitochondrial DNA has been shown to induce neutrophil infiltration into brain tissue via activation of the STING-NLRP3/IL-1β axis, worsening AD progression. ⁷⁰

Our keyword emergence analysis reveals synaptic loss, gut microbiota, and inflammasome activation as the primary research hotspots in the pathophysiological mechanisms of AD based on mouse models. These emerging themes reflect the shift towards a more systematic and multidimensional approach in AD pathogenesis research, encompassing areas such as synaptic integrity, microbiota interactions, and inflammatory responses. Future studies should continue to delve deeper into these mechanisms, with an emphasis on developing novel therapeutic strategies and biomarkers to address the underlying pathological processes of AD.