Relevant data was searched and obtained from the WoSCC in 2008–2022. Literature retrieval was conducted within a day (4 September 2022) to avoid fluctuations in rapid updates of publications. The search formula was constructed to TS= (“Neurodegenerative” OR “Neurologic Degenerative” OR “Nervous System Degenerative” OR “Neurodegenerative Disorder” OR “Neurologic Degenerative Condition” OR “Neurologic Degenerative” OR “Degenerative Neurologic Disorder” OR Degenerative OR “Alzheimer Disease” OR “Cognitive dysfunction” OR “Brain” OR “Alzheimer’s disease” OR Alzheimer OR AD OR “Parkinson’s disease” OR PD OR “Huntington disease” OR “Huntington’s disease” OR HD OR “Amyotrophic lateral sclerosis” OR ALS OR “Spinal cord cerebellar ataxia” OR SCA OR “multiple sclerosis*“) AND TS= (“network pharmacology” OR “systems pharmacology” OR “network medicine”). A total of 752 articles were retrieved, the main types of literatures chosen for our study were articles and reviews, and the language was confined to English. Two researchers (ZJ and HSY) independently performed the original literatures search and identify any possible errors and discrepancies. Besides, they read the titles, abstracts, and even the full texts of paper to filter out papers that are closely related to our subject. Only 420 publications were enrolled. The detailed retrieval process was shown in Figure 1.

In this study, visual analysis was constructed using Microsoft Office Excel 2019, HistCite (version 2009.08.24), VOSviewer (version1.6.18), CiteSpace (version6.1), and Bibliometrix 4.1.0 Packages based on the R language.

Microsoft Excel was employed to make the tables and display the information of countries, authors, institutions and references.

HistCite, a document indexing and analysis software designed to process information from literature searches, helps scholars master the historical development of a field and identifies highly cited research and scholars (Garfield et al., 2006). In present study, it was used in calculating the publications, reference citation, total local citation score (TLCS) and total global citation score (TGCS). TLCS refers to total citations in the collection, which can reveal the popularity of the article in this field and is usually less than TGCS. And TGCS refers to total citations in Web of Science to papers in the collection. Although TGCS does not directly reveal the focus of the field, it can still help to find the most influential work in all disciplines.

Vosviewer, a software for building a visual bibliometric network based on network data (van Eck and Waltman, 2010), is hired to construct visual network atlas for systematically understanding the structure and dynamic development of scientific researches. In this study, VOSviewer 1.6.18 was used to visualize the cooperative networks of productive countries, journals and authors based on bibliographic data.

CiteSpace is a citation analysis software based on scientometrics, data, and information visualization to visualize the distribution, patterns, and relationships of scientific knowledge (Chen, 2004). It can build visual networks, compute intermediary centrality, and perform burst detection to reveal changes in emerging trends, identify research frontiers (Chen et al., 2012), and tag key keywords (Chen, 2006). The visualization map is composed of nodes and lines (Chen and Song, 2019). Nodes are sized based on the number of items, and connections between nodes indicate co-occurrences, collaborations, or citations. The betweenness centrality measure quantifies the importance and connectivity of a node’s position within the network. The higher the betweenness centrality it is, the more connections it passes through (Freeman, 1978). Circles with a betweenness centrality over 0.1 were highlighted in purple. In clustering, the modularity Q and mean silhouette value are the most important parameters. Clusters were considered dominant and persuasive when Q > 0.3 and mean silhouette value >0.5 (Xu et al., 2022). In present study, CiteSpace was employed to analyze co-cited authors and documents. CiteSpace can also identify burst citations and keywords. Basic parameters of CiteSpace were set according to our previous study (Dong et al., 2021).

The bibliometrix package in R 4.1.0 was used to automate the transformation and analysis of bibliographic information for selected publications, including institutional and periodical outputs, as well as author impact indicators. Indicators including number of publications and number of citations were used to evaluate the impact factors of the authors. The h-index commonly used to assess the level of academic output of a researcher (Bertoli-Barsotti and Lando, 2017), higher h-indix indicates higher scholarly impact (Hirsch, 2005). As a derivative index of h-index, g-index can further measure the academic impact and achievements of scholars (Ali, 2021).

A total of 420 English articles related to the network pharmacology in NDs were included. Figure 2A revealed an overall growing trend for annual and cumulative quantity of publications released from 2008 to 2022. Figure 2B showed that annual number of TGCS peaked in 2014 (n = 1934), and TLCS peaked in 2018 (n = 31). 2014–2022 was a period of stable development in this field with 394 publications. However, TGCS and TLCS fluctuated during this period. The above results showed that although the number of citations fluctuated, the overall volume continued to grow sharply.

From 2008 to 2022, a total of 53 countries participated in the research on network pharmacology in NDs. The United States had close connections with European countries (Figure 3A). China was the most prolific country (256 publications), followed by the United States (109 publications) and the UK (26 publications) (Figure 3B). There was a rising trend in the number of publications in the main distribution countries (Figure 3C). In Figure 3D, nodes are sized according to the number of publications, the lines indicate the cooperation, and thickness of line represents the strength of connections. China had the highest quantity of publications, while the United States had more links between other countries. Additionally, The US, Italy and China had the highest centrality over 0.1 (Table 1), indicating that these countries were the major research centers in the area of the network pharmacology in NDs. According to Figure 3E, although China had a great number of articles, it had a low proportion of co-operation with authors from other countries. The most cited countries were China (3,525), followed by United States (1,760) and Italy (801) (Figure 3F). The results mentioned above pointed out that China and the United States had significant contributions and influence in this field.

Information visualization helps to identify influential institutions and clarify their cooperation. Chinese institutions are the main driving force in this area and top three institutions ranked by centrality were China Acad Chinese Med Sci (0.06), Guangzhou Univ Chinese Med (0.04) and Beijing Univ Chinese Med (0.03) (Figure 4A; Table 2). However, their centrality were less than 0.1, indicating that there are no key research center and sufficient inter-agency cooperation among the top 10 institutions. Beijing Univ Chinese Med (24 publications), China Acad Chinese Med Sci (17 publications), Guangzhou Univ Chinese Med (16 publications) were the top three highly productive institutions in this field (Figure 4B; Table 2). Although Chinese Acad Sci was not the institution with the largest number of publications, it had the highest TGCS, representing its great influences in all disciplines. In Figures 4C, D, the institutions owning five or more publications were used to create collaboration network. Inter-agency collaboration is considered crucial motivation to conduct large-scale research. We observed that the cooperative relationship between institutions was relatively simple, and correlation was not close. In short, China is the main driving force in this filed, but interagency cooperation still needs to be strengthened.

A total of 2,511 authors contributed to researches on network pharmacology in NDs. Geerts H, from the United States, had the largest number of publications (25), followed by Fang JS (12) and Spiros A (12) (Table 3). Moreover, most of the influential authors were from China and the United States. As shown in Figure 5A, Geerts H cooperated closely with Roberts P, Carr R and Spiros A. In addition, Fang JS frequently collaborated with Wang Q. In Figures 5B–D, Geerts H had the largest number of publications, highest H-index and G-index, indicating his outstanding contribution to this field. As early entrants into the field, Geerts H and Spiros A continued to publish related papers from 2012 to 2021, and Geerts H almost had important outputs each year (Figure 5E). As noted in Figure 5F, Hopkins AL had the highest co-citations, followed by Ru JL and Shannon P, with centrality more than 0.1 marked with purple rings, suggesting that they had strong academic influences in this field.

All articles are distributed in 193 journals, and four of these journals have more than 10 papers. We derived impact factors (IF) and journal quartiles from the 2021 Journal Citation Reports. The top three high-yield journals were Frontiers in Pharmacology (IF5.988), Evidence-based Complementary and Alternative Medicine (IF2.650), and Journal of Ethnopharmacology (IF5.195). The top 10 most productive journals in Table 4 were mainly ranked in Q1 or Q2, indicating that the above journals had high scholastic reputations in this field. The three most frequently cited journals were Nucleic Acids Research (IF19.160), PLoS One (IF3.752) and Journal of Ethnopharmacology (IF5.195) ( Figure 6A ). The distribution of scientific journal issues is shown in a double map overlay (Figure 6B). The double map overlay of journals displayed one core citation way, the citation journals are mainly distributed in the area of molecular, biology and immunology, and the cited articles published in journals were predominantly contained in the area of molecular, biology and genetics.

Table 5 lists the highly cited articles, the top cited article was “TCMSP: a database of systems pharmacology for drug discovery from herbal medicines” by Ru JL in 2014 (94 citations), followed by Shannon P in 2003 (73 citations) and Hopkins AL in 2008 (68 citations). We built a visualization network of co-cited references (Figure 7A). “Wishart (2018),” “Szklarczyk D (2019),” and “Zhou YY (2019)” were frequently co-cited. Cluster analysis divided most of the relevant terms related to network pharmacology in NDs into 12 main categories (Table 6). The top 5 clusters were “network pharmacology,” “alzheimers disease,” “network medicine,” “systems biology,” and “quantitative systems pharmacology”. The silhouette value (S) represents the mean contour value of a cluster. Generally, if S > 0.5, cluster is considered proper, and if S > 0.7, cluster is persuasive. In this research, the S values of the top 12 clusters were more than 0.7, indicating that clustering was considered credible. Among all clusters, “network pharmacology” (Cluster 0#) was the largest cluster containing 68 articles (Table 6), which mainly discussed the application and the development of network pharmacology. In addition, references with the strong citations burst were mainly concentrated in the three clusters of 0#, 1#, 2# (Figure 7B), they also were the most recently formed clusters, suggesting that network pharmacology and alzheimers disease may be the latest hotspots.

Furthermore, we build a timeline for clusters in Figure 7C. Relatively, the application of network pharmacology and traditional Chinese medicine therapy of AD are newly concerned by researchers. Figure 7D lists the 11 references with the strongest citation explosions from 2010 to 2022. The most cited article was from Ru JL in 2014, which also had the largest number of citations. Based on the burst time, Barabasi AL in 2011 and Roberts PD in 2012 were most prominent. Barabasi AL et al. through the way of network pharmacology, systematically explored diseases’ pathogenic factors, recognition pathways and molecular relationships between different phenotypes to identify drug targets and biomarkers for specific diseases (Barabasi et al., 2011). Roberts PD et al. explored the potential of quantitative systems pharmacology in assessing drug activity and mechanism by analyzing the differential efficacy of memantine in different stages of Alzheimer’s disease (Roberts et al., 2012). The above results hinted that exploration and application of network pharmacology were widely concerned for a long time.

According to visual analysis of keywords by CiteSpace (Figure 8A), network pharmacology and Alzheimer’s disease were the two most prominent words. As shown in Table 7, the most common keywords were: network pharmacology, Alzheimer’s disease, brain, systems pharmacology and traditional Chinese medicine. Among them, double blind (0.24), inhibition (0.23), and amyloid precursor protein (0.2) showed highest intermediary centrality. The intermediary centrality represents the connectivity of a node, and the higher it is, the more terms can be connected through the node, indicating the importance of a node in the network.

Burst keywords detection can reveal the evolution of the topic, predict the development directions and excavate potential flash point. As shown in Figure 7B, the keywords were mainly concentrated between 2012 and 2020. Besides, the burst keywords with strong strength were “systems pharmacology,” “drug discovery,” and “discovery”, indicating that the application of systems pharmacology and the search for drugs during this period may have been a major concern for researchers. Furthermore, “amyloid precursor protein” had been followed for the longest time, which was an important molecule in the pathogenesis of AD. In keyword burst detection, “systematic pharmacology” and “drug discovery” have highest intensity, “amyloid precursor protein” had longest sustained burst, and the latest bursts were concentrated in “systematic pharmacology,” “database,” “traditional Chinese medicine,” and “alzheimers disease” suggested that the current hotspots may be focused on the exploration of therapeutic methods for AD. Burst detection in this research contained major types, molecular mechanisms, and hot research methods of neurodegenerative diseases.

This study visualized the relationships between network pharmacology and NDs to understand their trends and hotspots through the use of bibliometrics. However, there are a few shortcomings to this study. For instance, due to the limitations of current scientific measurement software, it is difficult to analyze data from multiple databases simultaneously, so this study only utilized the WOSCC database. Moreover, this study only included articles and reviews, not abstracts, conferences, or books. Therefore, more comprehensive studies and databases could be undertaken in the future for more complete and accurate analysis.