Mapping research trends of uveal melanoma: a bibliometric analysis

Background To explore the research trends for uveal melanoma with bibliometric methods using Web of Science Core Collection (WoSCC) and PubMed (PM). Methods To find UM-related studies, “uveal melanoma” was used as search term in the WoSCC and PM for the period time from 2000 to 2020. Bibliographic coupling analysis was used to investigate the journals with the highest number of UM-related publications. VOSviewer (VV) was used for mapping the knowledge domain and visualizing the co-occurrence of terms, authors, organizations, countries, co-citation literature, and keywords. The knowledge map based on WoSCC and PM was compared. Results In the WoSCC 3,748 articles were found, while in PM the search resulted in 3,403 articles. The number of original articles has steadily grown in general in the past two decades. The top ten authors were contributing to 23% (n = 856) of all publications, while the top 10 institutions published 41% (n = 1524) of all articles. The top 3 journals with the highest number of publications for UM-related research included Investigative ophthalmology & visual science, Ophthalmology, and British Journal of Ophthalmology. Co-occurrence analysis based on author keywords showed 6 clusters. The most frequent keywords included are metastasis, prognosis, and brachytherapy. The latest research hotspots focused on BAP1, immunotherapy and GNAQ. Conclusions Genetics and immunology are the latest research frontiers in uveal melanoma. There is a clear need for interdisciplinary, molecular and clinical research approaches to improve the fatal prognosis of uveal melanoma patients.


Background
Uveal melanoma (UM) is a common primary intraocular malignancy in Caucasians [1]. 85% of ocular melanomas are UM [2]. Unfortunately, this common malignant tumor has a poor prognosis and low survival rate. Metastasis to the liver is common, and half of posterior UM patients suffered from this within 15 years after the initial diagnosis and treatment [3]. Meanwhile, the 5-year all-cause mortality rate following enucleation for choroidal melanoma ranges from 16 to 53% [1]. Compared to common ophthalmic diseases, the number of studies and research on UM is limited. However, some research papers have been published in the last 20 years. Despite the relatively limited number of articles regarding UM research, researchers cannot read all of the UM literature, and abstracts, as a traditional type of literature, do not provide us with information on hot topics and current trends in the field of UM research.
Web of Science Core Collection (WoSCC) and PubMed (PM) provide established databases for literature search and bibliometric analysis, a method using mathematical statistics to analyze relevant literature, is now firmly established as scientific specialties and is an integral part of research evaluation methodology, especially within the scientific and applied fields [4]. With bibliometric analyses, it is possible to assess the growth in publications, international collaborations, author productivity, or keyword co-occurrence related to UM studies. Using visualization tools, a knowledge domain map through bibliometric hotspots and connections can be generated.
This study aims to explore the current research status of uveal melanoma related studies by mapping the knowledge domain with bibliometric methods using Web of Science Core Collection and PubMed, to identify research trends and to provide researchers a better overview and more information about UM related research over the last 20 years.

Methods
The Science Citation Index Expanded database in the Web of Science Core Collection (WoSCC) was included in this study. The searching keyword is Uveal melanoma, the document type was article, and the period was ''from 2000 to 2020.'' No language restrictions were set. Excel (2016; Microsoft) was used to create a database based on WoSCC and PubMed-generated data. VOSviewer [5] (1.6.15; Leiden University) was used to mapping the knowledge domain and visualizing the co-occurrence of terms, authors, organizations, countries, co-citation literature, and keywords.
For analysis modes, the bibliographic coupling analysis was performed for source journals and cooccurrence analysis for keywords. In Bibliographic coupling analysis, the relatedness of items is determined based on the number of references they share. In co-occurrence analysis, the relatedness of items is determined based on the number of documents they occur together.
For the knowledge map, the node size defines the number of published articles. The links between nodes represent relatedness. The color of a node determines the cluster it belongs to, and different clusters are represented by different colors.

Annual publications distribution
Based on WOSCC, the publication distribution was shown in Fig. 1. The number of average annual publications is 179.43 ± 68.90, and the average annual growth rate is 7.05%. The number of publications has steadily grown in general in the past two decades. The literature growth rate has been positive for the past 10 years, with peaks of growth in 2003, 2012, and 2015. The growth rate of 30% in 2013 was the highest within the last 20 years.

Research institutions and country and authors distribution
There are 2,650 institutions that published 3,748 articles. The top 10 institutions published 40.66% of all publications, which is 1,524 articles. The rank was organized in Table 1. In the last two decades, 3,768 matched articles originated from 78 countries. The top 5 countries included USA, Germany, UK, China, and the Netherlands. Due to the imbalanced institution distribution, co-authorship analysis was not performed. Authors and research institutions are highly correlated, and the distribution of research institutions is closely linked to the distribution of countries. A total of 13,979 authors contributed research to this area, with the top ten contributing 22.84% (n = 856) of all literature seen in Table 2.

Source journals analysis
Based on WoSCC, the bibliographic coupling analysis of source journals was performed. The top 10 journals were organized in Table 3. Investigative ophthalmology & visual science accepted most papers about UM (223, 5.95%); Ophthalmology accepted 114 articles (3.04%); British Journal of Ophthalmology accepted 126 (3.36%). The top 3 journals accepted 12.35% of articles in the past two decades. The bibliographic coupling analysis visualization is shown in Fig. 2. It is worth noting that in the distribution of journals, it is clear that there are 3 main blocks. The clusters are very close together, but the connections between different clusters are very far apart. In the knowledge map, greater distances indicate fewer connections. This distribution indicates that the research hotspots covered by individual journals are concentrated and limited, and it is not easy to find suitable journals for publication of cross-disciplinary research. The top three journals in the ranking are located in the first and second largest clusters, with no representative journals appearing in the third cluster. Additionally, according to total link strength, although Investigative ophthalmology & visual science includes more literature, Ophthalmology has a higher total link strength, meaning that it publishes articles that are more recognized by researchers as more influential.

Keywords distribution and analysis
According to co-occurrence analysis based on author keywords of WoSCC, the hotspots of keywords related to UM were found. We set a minimum number of occurrences of a keyword as 10. Of the 4286 keywords, 121 meet the threshold. Similar keywords will be clustered together, eventually forming six major clusters, marked in 6 colors (Fig. 3). By filtering or combining similar keywords, the following keywords were found to be more influential according to the total link strength: metastasis (308), prognosis (221), brachytherapy (192), BAP1 (154), immunotherapy (136), immunohistochemistry (134), survival (131), GNAQ (102), enucleation (93) and radiotherapy  The relatedness of items is determined based on the number of references they share. Total link strength attribute indicates the total strength of the bibliographic coupling links of a given sources with other sources (88). The top 10 keywords in every cluster were listed in Table.4. By analyzing the average year of first occurrence, among the research hotspots, the following are the highlights of research in recent 5 years: BAP1 (154), immunotherapy (136), GNAQ (102), GNA11 (72) and ipilimumab (51). They are mainly concentrated in green, purple and brown clusters. Overly visualization was shown in Fig. 4. Through the images, we can see that genetics and immunology are currently at the frontier of research and recognized by most researchers.
According to co-occurrence analysis based on author keywords of PubMed, we set a minimum number of occurrences of a keyword as 5. Of the 2501 keywords, 112 met the threshold. Similar keywords were clustered together, forming six major clusters, marked in 6 colors (Fig. 5). By filtering or combining similar keywords, the following keywords were found to be more influential according to the total link strength: metastasis (99), BAP1 (74), prognosis (59), immunotherapy (56) and brachytherapy (50). Due to the different formats of these two databases, even though they cover mostly the same literature, the knowledge maps they constitute are different. The knowledge maps constructed from these two databases showed that the hotspots were similar in UM studies.

Discussion
The development trend of research about uveal melanoma was shown by the quantity variation of academic papers. By analyzing the literature over the last 20 years, we found a steady growth of studies on UM in the last 5 years. As a malignancy, such growth rates reflect a more comprehensive understanding of the disease. And it is noteworthy that there are several time points when the discovery of new related genes might lead to a new wave of research. Although the mechanism of the disease is not fully understood, we do have a better understanding of the disease.
The most productive institutes are mainly located in the United States, the United Kingdom, the Netherlands, France, and Germany. Six of the top 10 research institutions are from Europe. Considering the highrisk population for UM, it is not difficult to infer that Europe and the United States are UM research centers. We are very grateful to these core authors for their contributions to the field, but the significance of the visual analysis results about the authors is not very obvious. These authors are studying this field for many years, so their findings have a longer exposure time. However, the authors' results can be used as a reference indicator when researchers choose their collaborators. Fig. 2 The bibliographic coupling analysis visualization of sources. The node size is determined by the number of published articles. In Bibliographic coupling analysis, the relatedness of items is determined based on the number of references they share. The links between nodes represent relatedness. The color of a node represents the cluster it belongs to, and different clusters are represented by different colors Through the bibliographic coupling analysis of source journals, we can determine the core journals in studies of UM. In Fig. 2, the apparent clustering of journals is determined. The results show a clear stratification of clinical and basic research and uneven distribution. This implicates the difficulties that basic research encounters in translating results into clinical applications. In the top 3 journals, Ophthalmology and British Journal of Ophthalmology belong to the same cluster, while Investigative ophthalmology & visual science belongs to the first cluster. Articles published in these journals are more likely to be cited. In other words, articles in these journals are more likely to disseminate their findings.
For the analysis of keywords, we used co-occurrence analysis. Through this analysis, we can see the hotspots that authors in this field are most concerned about and draw a knowledge background map. Therefore, we conducted a cluster analysis to explore the main topics of UM research. UM, keywords formed six main clusters and clustered together keywords with similar research topics (Fig. 3). Combining the characteristics of UM and the current status of UM research, the 6 clusters are analyzed as follows: Cluster 1 (Red): in this cluster, keywords are mainly related to traditional therapies. When treating uveal melanoma, the most important thing is to reduce mortality. However, patients and ophthalmologists are also committed to maintaining the visual function, cosmetic appearance, and quality of living [6,7]. Treating small to medium melanomas with Ru 106 was a success [8][9][10]. According to different studies, regular treatments like enucleation, brachytherapy, charged particle irradiation, and local resection have similar survival outcomes [11][12][13]. It is noteworthy that studies on traditional treatment methods still Fig. 3 The co-occurrence analysis visualization of keywords in WoSCC. The node size is determined by the number of published articles. In co-occurrence analysis, the relatedness of items is determined based on the number of documents they occur together. The links between nodes represent relatedness. The color of a node represents the cluster it belongs to, and different clusters are represented by different colors The more frequent of keyword according to the co-occurrence analysis, the higher rank it gets. The numbers after keywords show the total link strength of keyword according to the co-occurrence analysis accounted for the majority, but the studies were generally older. Cluster 2 (Green): genetically related prognosis of melanoma was focused in this cluster. GNAQ, a stimulatory a q subunit of heterotrimeric G-proteins, was found to be mutated in 40% of Ums [14]. Some studies showed that 83% of UM had somatic mutations in GNAQ or GNA11. Constitutive activation of the pathway involving these two genes appears to be a major contributor to the development of uveal melanoma [15]. These two genes caused a wave of research around 2010. Prior to this, around 2003, research in this area was focused on BRAF mutations. Although Mutations in the BRAF gene enhance the kinase activity have been described in [ 60% of cutaneous melanomas and premalignant melanocytic lesions, it is not common in primary uveal melanoma [16,17]. Within this cluster, several genetic research hotspots are beginning to emerge, they are strongly clustered in time, and the research is generally close to us.
Cluster 3 (Blue): metastasis of UM is most mentioned in this cluster. In this cluster, the research approach is more oriented toward oncology. Of greatest concern were UM metastasis, apoptosis, and invasion. Monosomy 3 as a significant predictor of both relapse-free and overall survival of UM [18] is in this cluster's leading position. In this cluster, however, there is no trend regarding specific genes or key nodes. This indicates that there are still many unclear mechanisms waiting to be studied in this area. Researchers can look for new research directions in this cluster. Cluster 4 (Orange): comparing with cluster 3, the research approach is more oriented toward ophthalmology. Within this cluster, the hotspots of research are the choroid, uvea, ciliary body. Immunohistochemistry, as a critical histopathological examination [19], appeals in this cluster. Also appearing is plaque radiotherapy, a common treatment method [20].
Cluster 5 (Purple): most keywords in this cluster mutation of BAP1 were noticed. Some specific genes such as BAP1, EIF1AX, and SF3B1 are related to uveal melanoma metastasis [21] and have prognostic value in UM. Many studies found loss of BAP1 in uveal melanoma metastasis may be mainly involved in the progression of uveal melanoma to an aggressive, metastatic phenotype [19,21]. A high prevalence of liver metastases, as a character of UM, is clustered here. Unlike clusters 2 and 3, there is an exact object of study, and the direction of study is essentially the same. Research on this gene is worth continuing and is expected to translate into clinical results.
Cluster 6 (Brown): Since immunotherapy has dramatically changed the treatment approach to cutaneous melanoma [22], it is no surprise that immunotherapy has become the hotspot in the latest cluster. The drugs currently in the spotlight are ipilimumab, pembrolizumab, and nivolumab [23][24][25]. The biomarker, which is closely related to immunotherapy, appears in this cluster at the same time. In this cluster, the literature's average publication date is relatively new, and the study objects are relatively specific. The co-occurrence analysis visualization of keywords in PubMed. The node size is determined by the number of published articles. In co-occurrence analysis, the relatedness of items is determined based on the number of documents they occur together. The links between nodes represent relatedness. The color of a node represents the cluster it belongs to, and different clusters are represented by different colors By comparing the WoSCC and PubMed databases, PubMed has fewer metadata to analyze, the aggregation effect is not apparent, and the connection is not strong enough to reveal the deep connection between individual studies in the field. This study's limitation is that the PubMed database tags are not uniform, resulting in lower quality of metadata acquisition compared to WoSCC. Therefore, the amount of literature covered in this study is not comprehensive.
In conclusion, the analysis of the UM literature for the past 20 years by using scientific data visualization tools permits researchers to find references for their potential research directions, to determine potential research collaborators, and to find proper journals for publishing their articles regarding UM. More importantly, it identifies current hotspots including immunotherapy, BAP1, and GNAQ, in UM research for the last two decades.
Since medical research's speed now exceeds the learning speed of general physicians and researchers, it is more important to use scientific tools to organize the unknown knowledge. Databases for standardized management of literature tags are more critical in the knowledge explosion era than knowledge itself. In addition, these results help clinicians to find the newest relevant literature and clinical information for state-of-the-art treatment concepts for UM.
Acknowledgements Not applicable.
Authors' contributions SL and LMH conceived and planned the experiments. SL, XH and JL collected data. YG, WF, SJ, PA WM and ACR contributed to the interpretation of the results. SL took the lead in writing the manuscript. All authors provided critical feedback and helped to shape the research, analysis and manuscript.
Funding Open Access funding enabled and organized by Projekt DEAL.
Availability of data and materials Not applicable.

Declarations
Conflict of interest The authors have no financial interest to disclose. Names of commercial manufacturers are provided for identification purposes only. There are no any conflicts of interest between the clinician (author group) evaluating the apparatus and the commercial entity selling it / making it available.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.