Analysis of co-cited references: a cluster of research and most cited papersCluster of research

We first constructed several cluster-based co-citation networks of retrieved references for 2000–2022, 2017–2022, and 2022, respectively. All three networks were validated to be well-structured and sufficiently credible (Q = 0.8329, S = 0.938 for the 2000–2022 network; Q = 0.7397, S = 0.9543 for the 2017–2022 network; and Q = 0.6672, S = 0.879 for the 2022 network, respectively). Each reference was represented by a single node, whose size was proportional to the times the reference has been co-cited. Detailed descriptions of the largest clusters of co-cited references were presented in Additional file 2: Table S1, and visualized networks of each identified cluster were illustrated in Additional file 1: Fig. S2. Moreover, we provided relevant information on the link walkthrough between clusters based on burst dynamics for the co-cited reference network (2000–2022) (Additional file 1: Fig. S3).

In the 2000–2022 co-citation network, we identified a total of 19 different clusters, each of which was assigned a cluster number that depends on their sizes (ranging from the largest size (#0) to the smallest size (#25)), along with other qualitative measures, including cluster label, cluster size (N), silhouette score (S), and mean year (Y) of co-cited references. To summarize how research topics in this field developed during the past 2 decades, we integrated these clusters into four major research trends: (a) epigenetic mechanisms of CVD, incorporating 9 clusters on “lncrna”, #0 (N = 223; S = 0.919; Y = 2015)”, on “microrna”, #2 (N = 198;S = 0.894; Y = 2005), on “circular rna”, #6 (N = 133; S = 0.983; Y = 201), on “myocardin”, #7 (N = 129; S = 0.981; Y = 2002), on fetal programming”, #10 (N = 105; S = 0.991; Y = 2005), on “dna methylation”, #11 (N = 80; S = 0.974; Y = 2015), on “homocysteine”, #13 (N = 69; S = 0.995; Y = 1999), on “histone modifications”, #14 (N = 57; S = 0.971; Y = 2008), and on “epitranscriptomics”, #15 (N = 38; S = 0.986; Y = 2018); (b) epigenetics-based therapies for CVD, incorporating 3 clusters on “exosomes”, #3 (N = 167; S = 0.947; Y = 2016), on “regeneration”, #9 (N = 120; S = 0.939; Y = 2011), and on “inclisiran”, #16 (N = 20; S = 0.998; Y = 2018); (c) epigenetic profiles of specific CVDs, incorporating 5 clusters on “atherosclerosis”, #1 (N = 209; S = 0.908; Y = 2011), on “hypertrophy”, #5 (N = 153; S = 0.907; Y = 2008), on “atrial fibrillation”, #8 (N = 122; S = 0.887; Y = 2012), on “cardiac fibrosis”, #12 (N = 72; S = 0.952; Y = 2018), on “cerebral ischemia”, #25 (N = 5; S = 1; Y = 2009); (d) epigenetic biomarkers for CVD diagnosis/prediction, incorporating only 2 clusters on “biomarkers”, #4 (N = 159; S = 0.923; Y = 2012) and on “risk factors”, #17 (N = 10; S = 0.995; Y = 2019) (Fig. 1).

Fig. 1

Co-citation network of references (2000–2022) with corresponding clusters obtained with CiteSpace. A Co-citation reference network with cluster visualization and citation bursts of hotspots. B Visualization map of the corresponding clusters and citation bursts of hotspots. Note: Each node represents a co-cited reference, with the size of the node being proportional to the number of times the reference has been co-cited. The tree rings surrounding the nodes refer to citation bursts of co-cited references

In addition, we analyzed the co-cited references published from 2017 to 2022 (Additional file 1: Figs. S4 and S5) with yearly time slices and those published in 2022 with monthly time slices (Additional file 1: Fig. S6), which permitted an overview of the major research trends that emerged more recently. Not surprisingly, we found considerable overlap in identified clusters between this network and the 1989–2022 network, and determined 4 clusters that appeared for the first time during this period: (a) cluster #6 on “mir-29a” (N = 59; S = 0.96; Y = 2013); (b) cluster #7 on “heart failure” (N = 56; S = 0.972; Y = 2014); (c) cluster #9 on “RNA methylation” (N = 24; S = 0.989; Y = 2018); (d) cluster #12 on “transdifferentiation” (N = 14; S = 0.992; Y = 2013). As for the 2022 network, several novel clusters on “ferroptosis”, #0 (N = 84; S = 0.811; Y = 2019), on “clonal hematopoiesis”, #3 (N = 63; S = 0.897; Y = 2018), on “ischemic heart disease”, #4 (N = 48; S = 0.86; Y = 2018), on “cardiomyocyte proliferation”, #5 (N = 43; S = 0.896; Y = 2019), on “single-cell technology”, #9 (N = 13; S = 0.97; Y = 2019), on “brca1”, #10 (N = 11; S = 0.995; Y = 2019), on “somatic cell reprogramming”, #11 (N = 10; S = 1; Y = 2018), on “angiotensinogen”, #12 (N = 10; S = 0.989; Y = 2019), on “twin-twin transfusion syndrome”, #13 (N = 8; S = 0.985; Y = 2018), and on “dex”, #14 (N = 5; S = 0.993; Y = 2018) was noted, representing the most attractive topics at present.

Most cited papers

We extracted the top 10 papers with the highest citation frequencies published during the period 2000–2022 (Table 1). Of all these papers, those ranking within the top 3 positions were David P. Bartel’s review on the current status of the knowledge of miRNA target recognition in mammals and the mechanisms whereby miRNA modulates the expression and activity of protein-coding genes (337 citations) [21], followed by Carè et al. work establishing the essential role of miR-133 in suppressing cardiac hypertrophy in vitro and in vivo (276 citations) [22], and Thum et al. research article that an upregulated cardiac fibroblast-specific miR-21 expression was observed in pressure overload-induced failing myocardium, and may further exert detrimental effects on the geometry and function of heart through activating extracellular signal-regulated kinase–mitogen-activated protein kinase (ERK–MAPK) signaling pathway (270 citations) [23].

Table 1 The top 10 most cited references

Moreover, we analyzed the impact of papers published during the period 2000–2022 and 2017–2021, respectively, using calculating the citation bursts (Additional file 3: Supplementary Tables S2S–V). The blue line is the timeline sliced year by year, and the red line is representative of how long a citation burst persists. The top 3 references with the latest and strongest beginning of citation bursts included “A circular RNA protects the heart from pathological hypertrophy and heart failure by targeting miR-223” published by Wang et al. in 2016 [24], “A long noncoding RNA protects the heart from pathological hypertrophy” published by Han et al. in 2014 [25], and “Circulating microRNAs: novel biomarkers and extracellular communicators in cardiovascular disease?” published by Creemers et al. in 2012 [26]. As for the last 5 years, the top 3 references were an updated report on the epidemiological statistics of CVDs in the USA launched by the American Heart Association (AHA) [27], Love et al. article that introduced a sophisticated R package, namely, DESeq2, for handling with count data produced in high-throughput sequencing assay, which has currently become one of the most widely used techniques in determining CVD-related epigenetic loci [28], and another research article demonstrating the utmost importance of a highly conserved lncRNA metastasis-associated lung adenocarcinoma transcript 1 (MALAT1) for enabling the normal function of vascular endothelial cells and stimulating angiogenesis [29].

Analysis of co-occurrence of keywords

The predominant goal of constructing co-occurrence networks of keywords is to gain a comprehensive overview of the current research status and to forecast the evolution of research hotspots over time. A single node in this network is indicative of a highly co-occurring keyword, with the node size depending on how frequently it occurs. The 2000–2022 and 2017–2022 networks both exhibited significant structure and adequate reasonability (Q = 0.303, S = 0.6868 for the 1989–2022 network; and Q = 0.3045, S = 0.6362 for the 2017–2022 network, respectively) (Fig. 2).

Fig. 2

Timeline visualization of networks of co-occurring keywords for the period 2000–2022 (A) and 2017–2022 (B). Note: Each node is representative of a co-occurring keyword, with the color of the node depending on the average publication year of all articles containing this keyword. A brighter node refers to a keyword that emerged more recently. The tree rings surrounding the nodes represent citation bursts of co-occurring keywords. The networks are weighted on total link strength across different nodes and scored on the average publication years. The identified cluster labels are marked in red, and placed on the right side of the networks

In the 2000–2022 network, we identified 9 distinct clusters: cluster #0, “oxidative stress”, cluster #1, “DNA methylation”, cluster #2, “heart failure”, cluster #3, “myocardial infarction”, cluster #4, “RNA interference”, cluster #5, “cardiovascular disease”, cluster #6, “congenital heart disease”, cluster #7, “extracellular matrix”, and cluster #14, “fatty acid oxidation”. When focusing on the last 5 years, we found several newly emerging clusters: cluster #2, “long non-coding RNA”, cluster #6, “the-beta 1”, cluster #7, “circulating miRNAs”, cluster #8, “cardiac arrest”, and cluster #9, “intermittent hypoxia”.

Keywords with the latest and highest citation bursts are highly predictive of hotspot topics, research frontiers, and future research trends. The top 3 keywords that appeared most recently and had the strongest beginning of citation bursts included “nlrp3 inflammasome”, “myocardial injury”, and “reperfusion injury” for the 2000–2022 network, while those for the 2017–2022 network were “machine learning”, “nlrp3 inflammasome”, and “risk prediction” (Additional file 3: Tables S2W–Z).

In addition, we employed VOSviewer to establish a network visualization of co-occurring keywords, in which the lines connecting pairs of keywords become thicker with increasing numbers of co-occurrences. An overlay visualization of co-occurring keywords—wherein one keyword was assigned a specific color that varies from blue to yellow depending on the mean years of publications of articles incorporating this keyword (keywords occurring in earlier years were colored in blue, and those appearing later were colored in yellow)—was also provided. Interestingly, despite five distinct clusters being screened out and marked with different colors, all these clusters were exceptionally similar in the evolving process of research trends, and an even distribution of newly emerging topics across the clusters was noted (Additional file 1: Fig. S7).

Publication outputs and major journals

We originally retrieved 30,180 articles regarding epigenetics in CVDs from the WoSSC database, ruled out 2418 non-relevant ones, and included the rest for further analyses (Additional file 1: Fig. S1). The yearly amounts of publications displayed an exponential rise since the early 2000s, and the last 5 years have witnessed a surge of publications, implying a rapidly expanding interest in this realm. During the same period, the average number of citations per year also tended to increase gradually, reached its peak in 2005, and then slightly declined but still maintained at a relatively high level; however, the entire course seemed to be more fluctuating (Additional file 1: Fig. S8).

From 2000 to 2022, the leading journal that published the most references was Scientific Reports, followed by the International Journal of Molecular Sciences and Circulation Research. For the last 5 years, the International Journal of Molecular Sciences, Scientific Reports, and Frontiers in Cardiovascular Medicine constituted the top 3 journals with the highest publications. Most journals displayed an approximately linear increase in the number of publications, while the International Journal of Molecular Sciences and Frontiers in Cardiovascular Medicine were the only two that experienced an inflection point during this period. The rising speed of the number of publications on the right side of the inflection point was higher compared to that on the other side of the inflection point, causing more rapid growth of publications in these two journals (Additional file 1: Fig. S9). An overlay visualization of the most cited journals over the past 5 years and a co-citation network of journals for the past 2 decades were also presented (Additional file 1: Fig. S10).

The top 3 journals with the latest and strongest beginning of citation bursts for the past 2 decades were Frontiers in Immunology, Frontiers in Genetics, and Bioscience Reports (Additional file 3: Tables S2G, H), while those for the last 5 years were Artificial Cells, Nanomedicine, and Biotechnology, Clinical Medicine Insights: Cardiology, and Brain and Behavior (Additional file 3: Tables S2I, J).

Analysis of cooperation network across countries and institutions

We constructed the co-citation networks of countries and institutions (Fig. 3), and listed the top countries and institutions ranked by number of citations and betweenness centrality (Additional file 6: Table S5). China made the most prominent contribution to citation counts (n = 9910), followed by the United States (n = 8271), and Germany (n = 1984). Spain ranked first in betweenness centrality (0.23), followed by England (0.12), and France (0.11). When focusing on the last 5 years, the top 3 positions in the rankings of citation numbers remained unchanged, whereas France began to surpass England and Spain and became the area with the highest degree of betweenness centrality.

Fig. 3

Co-citation network of co-authors’ countries (2000–2022) (A) and co-citation network of co-authors’ institutions (2017–2022) (B) with corresponding clusters (C)

The top 3 institutions ranked by citation numbers were Nanjing Medical University (n = 472), Harbin Medical University (n = 439), and Shanghai Jiao Tong University (n = 428). In contrast, those ranked by betweenness centrality were Harvard University (0.11), Baylor College of Medicine (0.06), and Columbia University (0.04). The top 3 most cited institutions identified within the last 5 years were quite similar, except Harvard Medical School. In terms of betweenness centrality, the University of Medical Center Utrecht (0.08) achieved the best performance during this period, followed by the University of Texas Health Science Center (0.05) and the University of Naples Federico II (0.05).

Moreover, the top 3 countries that possessed the latest and highest citation bursts over the period 2000–2022 included Russia, Pakistan, and South Africa (Additional file 3: Tables S2A, B). As for the institutions, Central South University and Chinese Academy of Medical Sciences & Peking Union Medical College, and Shandong First Medical University retained leading positions for either the 2000–2022 period or the 2017–2022 period (Additional file 3: Tables S2C–F).

Analysis of co-authorship network

To analyze and visualize the collaboration between different researchers based on the amount of co-authored publications, a well-structured and highly credible co-authorship network (Q = 0.7456, S = 0.8777) was thus established via CreateSpace (Fig. 4). In this network, we found the three most important clusters included: cluster #0, “autophagy”, cluster #1, “heart failure”, cluster #2, and “DNA methylation” (Additional file 4: Table S3). According to the findings of the burst analysis, the top 3 co-authors considered the most recent and influential contributors from 2000 to 2022 were Liu Y, Wang J, and Zhang J (Additional file 3: Tables S2K, L), and those for the 2017–2022 period were Katus HA, Wang Y, and Wang X (Additional file 3: Tables S2M, N). In addition, a similar co-authorship network was obtained with VOSviewer (Additional file 1: Fig. S11).

Fig. 4

Co-authorship network (A) with corresponding clusters (B) for the period 2000–2022

We further investigated the co-citation relations among authors from 2017 to 2022 (Additional file 1: Fig. S12). The top 3 co-cited authors who possessed the latest and strongest citation bursts were Bonauer A, Small EM, and Fichtlscherer S for the period 2000–2022 (Additional file 3: Tables S2O, P), and were Bolli R, Zhu HY, and Zhang XQ for the period 2017–2022 (Additional file 3: Tables S2Q, R, Additional files 5 and 6).

Bibliographic coupling analysis of countries, institutions, journals, references, and authors

Next, we employed VOSviewer to examine the bibliographic coupling of the publications in terms of different countries/institutions/authors/journals/references (Fig. 5), and calculated the total link strength in the bibliographic coupling networks to illuminate the relatedness of research domains (Additional file 7: Table S6). Among all countries, the United States had the strongest total link, followed by China and Germany. As for institutions, Hannover Medical School occupied the top position, followed by Harbin Medical University and Harvard University. Then, the top 3 journals that possessed the largest total link strength included International Journal of Molecular Sciences, Plos One, and Circulation Research, and the top 3 references were: “Non-coding RNAs in Development and Disease: Background, Mechanisms, and Therapeutic Approaches” published by Beermann et al. in 2016 [30], “MicroRNA regulatory networks in cardiovascular development” published by Liu et al. in 2010 [31], and “MicroRNAs add a new dimension to cardiovascular disease” published by Small et al. in 2010 [32]. Finally, Wang Y, Zhang Y, and Li Y constituted the three authors that performed best in total link strength.

Fig. 5

Bibliographic coupling networks of countries (A), institutions (B), journals (C), references (D), and authors (E) (weighted on the total link strength). Note: The minimum number of documents in a country should exceed 583; the Minimum number of citations of a document should exceed 240,490; the Minimum number of documents of a journal should exceed 10,519; Minimum number of documents of an author should exceed 30,579; Minimum number of documents of an institution should exceed 25,507