This article is mentioned in:

Introduction

The incidence of tumours has risen over the past half-century, resulting in cancer becoming one of the most lethal human diseases. This threat to humanity is great and continues to increase, and the urgency with which humans address this challenge is also increasing. Rapid developments in science technology over the past century have resulted in an independent medical discipline, oncology, which has been further divided into a number of branches.

With the rapid development of the oncology field, several questions have been raised: What are the primary research centres conducting research in this field? What are the major topics of current oncology research? Which are the most important academic communities? The answers to these questions are essential for developing effective measures to address the updated status of oncology research and practices. We used two information techniques, document overview and knowledge domain visualisation (KDViz) (1,2), to answer the above questions and create a comprehensive picture of the field.

To visualise the intellectual structure of scientific fields, two main scientometric methods can be employed: document co-citation analysis and co-authorship analysis (3–8). Document co-citation analysis has been used in various theoretical and empirical studies, including research concerning anaesthesia (3), environmental science (4), knowledge management (5), ubiquitous computing (6), mapping scientometrics (7) and stem cells (8). Co-authorship analysis has been applied to various research fields, including those involving digital libraries (2) and hypertext (7). Chen and Liu (2) revealed the co-authorship pattern of scientometrics using the data from Science Citation Index (SCI). These two methods are used to explore research fronts and hotspots and the relationships between collaborators in selected scientific fields. These above studies demonstrate their practical value and advantages over document overview, but they are rarely used in medical research.

In the present study, we used KDViz for co-citation and co-authorship analysis to reveal the main research centres, topics and academic communities of oncology to probe the intellectual structure of oncology. Our study used a novel method and a unique perspective. We expect that it will reveal an overview of the development in this field from 2001 to 2010. The research fronts are identified based on indicators computed by CiteSpace without the intervention of domain experts or prior working knowledge of the topic. This approach makes the analysis repeatable with new data and verifiable by different analysts.

Materials and methods

Materials

Based on 185 oncology journals listed in the 2010 annual JCR report from the Web of Science database, we chose the top 10% by impact factor (IF) [19 journals (A)] and total citations [19 journals (B)]; 8 journals were in both groups A and B. We assigned 30 journals to be our data.

After retrieving 30 journals from the SCI database with the literature format ‘article’ and ‘review’, and removing those with the format ‘news’, ‘meetings abstract’, ‘letter’ and other non-original studies, we named the files ‘download*. txt’ and downloaded them in their fully recorded format with references. In total, we retrieved 143,152 articles from 2001 to 2010, including the authors, title, keywords, abstract and citations as sample data for our study. The last date on which we retrieved data was November 1, 2011.

Methods

Knowledge domain visualisation

KDViz is a computer-supported information processing technology that can reveal the developmental process and structure of scientific knowledge in graphical form by analysing information such as authors and publications and defining their relationships. These relationships are expressed in two- or three-dimensional knowledge landscapes to effectively describe large amounts of data and outline the structure and evolution of a scientific field (1,2).

Chen from Drexel University designed the visualisation software CiteSpace II, which is written in Java and can be used to analyse multiple-perspective co-citation networks to identify and display new trends in scientific development based on a large number of scientific studies (9). Knowledge maps drawn by CiteSpace II are not only able to predict future trends in research but can also aid in understanding the current forefronts (6,9–11). Chen also used CiteSpace II to draw knowledge maps on large-scale biological populations that are now extinct (1981–2004), terrorism (1990–2003)(12) and emerging trends in regenerative medicine (13).

In this study, CiteSpace II was used for the co-citation analysis of network graphs (Figs. 1 and 3).

Figure 1 Mapping core country/institution distribution of oncology research, 2001–2010. This image displays the number and the time of studies published by the countries/institutions as the form of a ‘ring’ visualization, and reveals where the cooperation among countries/institutions lies, as well as the intensity of cooperation. The annual ring of citation is based on the country’s/institution’s citation history. The color of the ring represents the corresponding time of the quotation, and the thickness of the ring is directly proportional to the frequency of citations in a time section. The purple node in the middle of the annual ring represents the impact and the importance of a country/institution. The larger the node and the more purple it appears, the greater is the importance of the country/institution. The connection line between two nodes indicates co-cited relations and co-citation intensity. The higher the frequency of the co-citation, the thicker the connection line is and the closer the nodes are. The color of the connection line represents the first co-citation year (15). (A) Mapping core country distribution of oncology research, 2001–2011; (B) mapping core institution distribution of oncology research, 2001–2010.

Figure 3 Mapping of oncology research fronts, 2001–2010. One annual citation ring represents one literature. The larger the annual citation ring is, the higher the frequency of citations in the literature. The connecting lines among rings show co-cited relations and co-citation intensity. Different background colors represent different automatic clusters by Citespace II. For each cluster, the more connection lines, the higher is the frequency of co-citation events, and the more important is the cluster (15). (a) The automatic clustering and function of the term identity of Citespace II shows four major oncology research front domains which are divided into 55 branches (removal of the clusters of the same name) based on the literature aggregation and group influence. (b) The most influential literature for each major frontier domain without background color.

Co-citation analysis

Co-citation analysis is the most influential citation analysis method and can be used not only to reveal the developmental status and changes in the structure of a scientific field but also to study the research fronts and domains and provide support for those making critical decisions in the science and technology field. Co-citation analyses include document co-citation analysis, author co-citation analysis and institution co-citation analysis. In this study, we introduced document co-citation analysis, and the principles of the aforementioned analyses are similar.

Document co-citation analysis (DCA) studies a network of co-cited references (14). The fundamental assumption is that co-citation clusters reveal the underlying intellectual structure. The notion that cited documents are concept symbols was introduced by Small (14). He found a high degree of uniformity in how specific concepts and specific references to documents (cited documents) are associated in the chemistry literature. These cited documents serve as symbols for scientific ideas, methods and experiments. The idea is further extended to clusters of noun phrases that are extracted from document citations. From the concept symbol perspective, the study of a co-citation network focuses on interpreting the nature of a cluster of cited documents and the interrelationships between the clusters (10). The principle of document co-citation analysis is as follows. If two documents are cited together in one or more articles, we say that the two documents are co-cited. Higher co-citation frequencies indicate closer links between the documents. Thus, based on the document citation relationship, we can analyse the affiliations between documents. If the documents are divided into clusters and classes, we can analyse the fronts of the current research according to the contents of documents (15).

Social network analysis (SNA)

SNA focuses on ties between social entities; it involves the mapping and measurement of relationships between components in a system (16). Graphs are used to detect and interpret patterns of social ties. The vertices in the graphs represent social actors, and the edges represent the social interactions between the actors. This representation allows us to apply graph theory, a branch of mathematics, to the analysis of what would otherwise be an inherently elusive and poorly understood problem: the tangled web of our social interactions. In the present study, SNA helps reveal the connection between authors in the oncology field between 2001 and 2010.

The co-authorship network is an important type of social network and has been widely used to detect the structure of scientific collaborations and status of individual researchers. Analysing collaboration through co-authorship is advantageous as it is inexpensive and practical (17). In the past few years, many studies have shown that the rates of research collaboration, as measured by co-authorship, have greatly increased. Gossart and Ozman (18) analysed the collaboration patterns of Turkish social sciences and humanities (SSH) by examining co-authored papers. Lu and Feng (19) proposed a new method called ‘extensity centrality’ to analyse the importance of authors in co-authorship networks. Yu et al(20) analysed research groups from the co-authorship network in the oncology field in China. Morel et al(21) used co-authorship network analysis to generate valuable information on the strategic planning, implementation and monitoring of the program launched by the Brazilian Government to fund scientific research.

Results and Discussion

Data were analysed to generate the rich visual maps shown in Figs. 1–3.

Primary research bodies

Core countries

Citation tree rings represent the citation history of a country/institution. The colour of a citation ring denotes the time slice of corresponding citations. The thickness of a ring is proportional to the number of citations in a given time slice. Larger rings indicate a higher frequency of citations for the publications. Fig. 1A shows the top 20 most-cited countries, which are also listed in Table I.

Table I The most-cited countries and their citation frequencies in oncology (top 20, 2001–2010).

Table I

The most-cited countries and their citation frequencies in oncology (top 20, 2001–2010).

Serial no.	Country	Citation frequency
1	USA	16,463
2	Japan	1809
3	Germany	1756
4	France	1523
5	England	1454
6	Canada	1334
7	Italy	1326
8	The Netherlands	719
9	P.R. China	583
10	Spain	507
11	Australia	499
12	Sweden	415
13	Republic of Korea	387
14	Switzerland	337
15	Israel	283
16	Belgium	277
17	Taiwan	276
18	Scotland	255
19	Austria	203
20	Denmark	201

In Fig. 1A, the citation frequency of the United States is 16,463, which accounts for 53.8% of the total citation frequency of the top 20 countries. The United States is not only the most-cited country, but its ring is evenly distributed, indicating that the United States has always been the leader of the international field of oncology research. The sole developing country in the top 10 is China, which ranks 9th. We did not find connection lines between the United States and China among the top 20 countries. Therefore, we can say that they are not co-cited, indicating that they have distinct research directions.

Core institutions

Fig. 1B shows that the contemporary oncology studies are primarily from the University of Texas, the NCI, Harvard University and the Memorial Sloan Kettering Cancer Center. Regarding both the quantity of issued studies and the citation frequency, the University of Texas ranks 1st, the NCI ranks 2nd and Harvard University ranks 3th. The rings of the NCI, Harvard University and other institutions are basically evenly distributed, indicating that their research strengths continued to steadily increase over the past 10 years. Fig. 1B shows the top 20 most-cited institutions and their citation frequency, which are also listed in Table II.

Table II The most cited institutions and their citation frequencies in oncology (top 20, 2001–2010).

Table II

The most cited institutions and their citation frequencies in oncology (top 20, 2001–2010).

Serial no.	Institution	Citation frequency
1	University of Texas	1177
2	National Cancer Institute	1007
3	Harvard University	802
4	Memorial Sloan-Kettering Cancer Center	555
5	University of California, San Francisco	368
6	Dana-Farber Cancer Institute	365
7	University of Pennsylvania	362
8	University of Michigan	353
9	Johns Hopkins University	350
10	University of California, Los Angeles	324
11	University of Pittsburgh	317
12	The University of Texas: MD Anderson Cancer Center	312
13	Mayo Clinic	287
14	Baylor College of Medicine	285
15	University of Toronto	273
16	University of Tokyo	271
17	Ohio State University	262
18	Vanderbilt University	258
19	Duke University	249
20	Brigham and Women’s Hospital	236

Core academic communities

We performed co-authorship analysis using SNA and Pajek was to develop the network graphs (Fig. 2).

Figure 2 The 36 cohesive oncology co-authorship network, 2001–2010. One vertex represents one author. The vertex size is proportional to the productivity of the author, while the thickness of the lines indicates the strength of connection between two authors. The thicker the line between the two vertices, the closer the relationship is. The value between two vertices represents the frequency of cooperation instances.

As shown in Fig. 2, our dataset was composed of 118,871 authors and was submitted to Pajek. To obtain a clear picture, we deleted the lines with values less than 20, i.e., we removed the authors whose cooperation instances were below 20. As shown in Fig. 2, the most collaborative authors were divided into 36 relatively dense components, including 91 vertices, showing that there were 36 close collaborative communities performing oncology research composed of 91 authors (Table III). The largest vertex was found in the white cluster and represents the author with the highest productivity, Nakamura Y, who published 117 studies from 2001 to 2010. The largest academic community (the biggest component) is the red cluster in the middle, which contains 8 authors who are mainly from the Harvard University School of Medicine, the Division of Hematology and Oncology of the Dana Farber Cancer Institute, and the Jerome Lipper Multiple Myeloma Center of Boston. Their research is focused on multiple myeloma. The 5 authors in the second component are all affiliated with the Department of Neurosurgery of the University of Illinois College of Medicine at Peoria, and their research focuses primarily on angiogenesis. The 4 authors in the third component are all from the Department of Leukemia of the MD Anderson Cancer Center at the University of Texas and perform research primarily on acute lymphocytic leukaemia (ALL). Furthermore, two vertices connect in the peripheral lap, which indicates that two authors in these components have cooperated at least 20 times; we consider these laboratories to be cohesive subgroups.

Table III Components for the most collaborative authors in Oncology Research (2001–2010).

Table III

Components for the most collaborative authors in Oncology Research (2001–2010).

Cluster	Freq	Freq %	CumFreq	CumFreq %	Representative
1	8	8.7912	12	13.1868	Hideshima T
2	5	5.4945	58	63.7363	Rao JS
3	4	4.3956	18	19.7802	Gazdar AF
4	4	4.3956	22	24.1758	Tsunoda T
5	4	4.3956	75	82.4176	Faderl S
6	3	3.2967	53	58.2418	Lee JJ
7	3	3.2967	63	69.2308	Sarkar D
8	3	3.2967	80	87.9121	Rajkumar SV
9	3	3.2967	87	95.6044	Kucia M
10	2	2.1978	2	2.1978	Baylin SB
11	2	2.1978	4	4.3956	Calin GA
12	2	2.1978	14	15.3846	Schellens JHM
13	2	2.1978	24	26.3736	Andreeff M
14	2	2.1978	26	28.5714	Janne PA
15	2	2.1978	28	30.7692	Spitz MR
16	2	2.1978	30	32.967	Wang LE
17	2	2.1978	32	35.1648	Fusco A
18	2	2.1978	34	37.3626	Grant S
19	2	2.1978	36	39.5604	Kinzler KW
20	2	2.1978	38	41.7582	Rimm DL
21	2	2.1978	40	43.956	Maher ER
22	2	2.1978	42	46.1538	Sidransky D
23	2	2.1978	44	48.3516	Inoue H
24	2	2.1978	46	50.5495	Henshall SM
25	2	2.1978	48	52.7473	Valero V
26	2	2.1978	50	54.9451	Singh RP
27	2	2.1978	60	65.9341	Debatin KM
28	2	2.1978	65	71.4286	Ellis LM
29	2	2.1978	67	73.6264	Sparreboom A
30	2	2.1978	69	75.8242	McCubrey JA
31	2	2.1978	71	78.022	Imoto I
32	2	2.1978	77	84.6154	Gertz MA
33	2	2.1978	82	90.1099	Armstrong L
34	2	2.1978	84	92.3077	Hochhaus A
35	2	2.1978	89	97.8022	Tefferi A
36	2	2.1978	91	100	Van Der Velden VHJ

Therefore, all of the aforementioned authors and institutions have played an important role in forming and connecting the collaborative oncology research network. The oncology research strength is basically balanced between the communities over the 10 years studied in this study. Only one academic community had greater impact and features (the largest red cluster) than the other communities. We also found that researchers preferred to collaborate within their own institution.

Oncology research fronts

Fig. 3 and Tables IV and V show the results of our document co-citation analysis using Citespace II.

Table IV The 4 largest DCA clusters of the oncology research network (2001–2010).

Table IV

The 4 largest DCA clusters of the oncology research network (2001–2010).

Field	C#	Size	Silhouette	Top terms by tf* idf	Top terms by LLR (P=0.0001)	Related clusters
A	#9	100	0.754	(7.04) kruppel-like factor; (16.49) CpG island methylation; (15.51) p63α; (15.51) p53 mutant: (15.45) microsatellite instability; (15.44) promoter methylation; (13.46) microRNA replacement therapy; (15.06) rassf1a; (14.36) human telomerase; (14.36) aberrant promoter methylation; (12.31) osteosarcoma U2OS cells; (10.78) p14; (9.73) Snail	(203.73) p53; (117.03) microsatellite instability; (103.63) p14; (97.86) rassf1a; (97.46) microRNA; (89.71) hypermethylation; (52.91) methylation; (27.41) snail	#6, #48, #49, #7, #14, #15, #16, #20, #50, #19
B	#31	69	0.822	(17.04) metastatic colorectal cancer; (16.98) metastatic breast cancer; (16.48) gefitinib; (15.67) endostatin; (14.4) adjuvant tamoxifen; (13.46) regional relapse; (11.16) microarray analysis; (10.62) Myc-induced microRNA; (10.62) platinum-doublet chemotherapy; (10.57) bevacizumab; (6.94) vascular endothelial growth factor	(94.55) tamoxifen; (76.03) gefitinib; (54.37) endostatin; (53.48) tumor growth; (49.74) bevacizumab; (36.21) metastatic colorectal cancer; (14.37) vascular endothelial growth factor	#29, #34, #44, #38, #36, #35, #37, #33, #30, #45
C	#23	45	0.346	(12.45) mitochondria; (12.14) death receptor; (11.87) adenine nucleotide translocator; (8.73) Bcl-2 protein interaction; (8.73) Bcl-2 family interaction; (11.42) tumour growth	(99.55) Trail; (94.12) factor-related apoptosis-inducing ligand; (41.98) drug-induced apoptosis; (41.09) adenine nucleotide translocator; (38.37) apoptosis; (25.41) Bcl-2 protein	#24, #25, #26, #21, #22, #17
D	#46	41	0	(16.49) vaccine; (15.86) dendritic cells; (14.94) multiple myeloma; (14.7) cyclooxygenase-2; (14.7) chemokine	(147.93) vaccine; (87.73) patient; (79.36) prostate cancer; (68.49) novel

Table V The most affective authors and their most-cited study in oncology research fronts (2001–2010).

Table V

The most affective authors and their most-cited study in oncology research fronts (2001–2010).

Serial no.	Research front	Representative author	The author’s most-cited study	Frequency
1	The mechanisms of abnormal oncogene expression	Herman JG	DNA methylation changes in hematologic malignancies: biologic and clinical implications (Leukemia, 1997) (23)	337
		Levine AJ	A single nucleotide polymorphism in the MDM2 promoter attenuates the p53 tumor suppressor pathway and accelerates tumor formation in humans (Cell, 2004) (29)	316
		He TC	PPAR delta is an APC-regulated target of non-steroidal anti-inflammatory drugs (Cell, 1999) (46)	218
		Serrano M	Cellular senescence in cancer and aging (Cell, 2007) (47)	208
		Jones PA	The fundamental role of epigenetic events in cancer (Nat Rev Genet, 2002) (24)	195
2	Tumor metastasis and angiogenesis	Therasse P	New guidelines to evaluate the response to treatment in solid Tumors (J Natl Cancer Inst, 2000) (48)	498
		Lynch TJ	Activating mutations in the epidermal growth factor receptor underlying responsiveness of non-small cell lung cancer to gefitinib (N Engl J Med, 2004) (34)	307
		Paez JG	EGFR mutations in lung cancer: correlation with clinical response to gefitinib therapy (Science, 2004) (33)	277
		Hurwitz H	Bevacizumab plus irinotecan, fluorouracil, and leucovorin for metastatic colorectal cancer (N Engl J Med, 2004) (31)	194
		Pao W	EGF receptor gene mutations are common in lung cancers from ‘never smokers’ and are associated with sensitivity of tumors to gefitinib and erlotinib (Proc Natl Acad Sci USA, 2004) (49)	141
3	The relationship between cancer cells and apoptosis	Green DR	The pathophysiology of mitochondrial cell death (Science, 2004) (39)	152
		Walczak H	Tumoricidal activity of tumor necrosis factor related apoptosis-inducing ligand in vivo (Nat Med, 1999) (41)	134
		Lip GY	Evidence of platelet activation in hypertension (J Hum Hypertens, 1997) (50)	134
		LI HL	Solution structure of BID, an intracellular amplifier of apoptotic signaling (Cell, 1999) (51)	134
		Ashkenazi A	Safety and antitumor activity of recombinant soluble Apo2 ligand (J Clin Invest, 1999) (42)	131
4	Tumor vaccines	Hanahan D	The hallmarks of cancer (Cell, 2000) (52)	695
		Davies H	Mutations of the BRAF gene in human cancer (Nature, 2002) (53)	251
		Jemal A	Cancer statistics, 2008 (CA Cancer J Clin 2008) (54)	195
		Lengauer C	14-3-3 Sigma is required to prevent mitotic catastrophe after DNA damage (Nature, 1999) (55)	185
		Sambrook J	Dominant negative ATM mutations in breast cancer families (J Natl Cancer Inst, 2002) (56)	179

Table IV shows the automatically chosen cluster labels of the 4 largest DCA clusters, along with their size, silhouette value and background colour shown in Fig. 3a. The top-ranked title terms according to LLR and tf* idf were selected as cluster labels. The largest cluster (‘CpG island methylation’, ‘p63’, ‘p53’ and ‘microsatellite instability’ (#9 and its related clusters) had 100 members and a reasonably high silhouette value of 0.754, which suggests a homogenous structure. The second largest cluster (#31 and its related clusters) had 69 members and a silhouette value of 0.822, which was the highest among the four co-citation networks, and was labelled ‘metastatic colorectal cancer’, ‘endostatin’ and ‘vascular endothelial growth factor’. The third largest cluster (#23 and its related clusters) had 45 members and was labelled ‘mitochondria’ and ‘death receptor’. The fourth largest cluster (#46) was labelled ‘vaccine’ and had the lowest silhouette value of 0, indicating a heterogeneous structure since there were no links found in this cluster (Fig. 3).

Other candidate labels for the clusters included ‘trastuzumab’ (#61) and ‘akt pathway activation’ (#57), confirming that these clusters are subclusters (Fig. 3).

As shown in Fig. 3, after removing a few small clusters that did not have relevant themes, we determined the top 5 representative scholars and their most influential studies in the four major research fronts (Table V).

Research front 1

In our interpretation of Fig. 3, we noted that the most important field of the four major fronts was Field A, which primarily involves the study of the mechanisms of abnormal oncogene expression. Field A includes the study of microsatellite instability (#7 and #10), DNA methylation (#9), the role of transcription factors in oncogene expression, tumour-suppressor genes (#48-#52) and telomerase (#15 and #16).

Microsatellite instability in a human colon cancer cell line was reported by Liu et al(22). They demonstrated that inhibiting the expression of the mismatch-repair (MMR) gene is caused by the inactivation of the two alleles of the gene. Their research is a milestone in revealing the mechanism of tumorigenesis.

Knudon’s two-hit model is now considered a mechanism of tumorigenesis, and some loci of microsatellite polymorphisms are used to examine microsatellite instability. In addition to microsatellite instability in autosomes, various researchers are focusing on the relationship between mitochondrial DNA microsatellite instability and cancer.

Another important area of research involves DNA methylation (#9). MLH1, one of the mismatch-repair genes, is often hypermethylated within its 5′ region in the apparently normal colonic epithelium of patients with microsatellite instability. In another example, during the progression of leukaemia, the hypermethylation state of several gene promoters was observed to be abnormal. This result suggests that epigenetic abnormalities could be used to monitor disease activity during therapy (23). Based on the opinion of Jones and Baylin (24), in neoplasia, the chromatin structure and DNA methylation patterns including the specific component differ from normal cells. These epigenetic changes, particularly regarding aberrant promoter hypermethylation, affect gene expression in tumour progression. Many clinical trials have found that DNA methylation plays an important role in the development of cervical cancer, ovarian cancer and endometrial cancer. Therefore, DNA methylation is a candidate epigenetic marker that can be used to diagnose tumours at an early stage and provide prognostic evaluation.

Various transcription factors are also involved in the abnormal expression of oncogenes. Kruppel-like transcription factors (#6) that contain zinc finger domains play a dual role in inhibiting or promoting tumour formation by regulating cell proliferation, cell differentiation, apoptosis and the expression of target genes. The transcription factor Snail, a member of the SNAIL superfamily (#18 and #19), is a zinc finger protein found in the Drosophila embryo in 1984 (25). Subsequently, Snail and its homologues have been found in other vertebrates, including humans. An increasing number of studies have shown that tumour invasion and metastasis are closely related to the overexpression of Snail. Thus, Snail can be used as a molecular marker for early diagnosis and prognosis in cancer. The investigation results by Tetsu and McCormick (26) (whose citation frequency ranked second in this field) showed that β-catenin activates transcription from the cyclin D1 promoter, and the promoter sequences related to consensus TCF/LEF-binding sites are necessary for activation.

Our map analysis showed that another small cluster was closely related to Field A, which involves several tumour-suppressor genes: p53, p63 and p14. The p53 tumour-suppressor protein plays a vital role in regulating cell growth following exposure to various stress stimuli. In normal cells, p53 induces either growth arrest, preventing the replication of damaged DNA, or apoptosis, which is important for eliminating defective cells (27). As a tumour suppressor, p53 is activated by a number of stressors to induce apoptosis and cell cycle arrest (28). In at least half of all cancers, the p53 gene is mutated (29). Many oncogenes belong to the same gene family in the human genome. The p53 gene family has become one of the most comprehensive and thorough focuses of gene family research. p63, a homologue of p53, is more complex than p53. In recent years, the search for a relationship between the abnormal expression of p63 and tumorigenesis has been a hotspot of tumour research. Another newly discovered tumour-suppressor gene, p14ARF, is an important member of the cell cycle network. The cell cycle is monitored by p14ARF via the p53-MDM2 and non-p53 pathway to inhibit tumour development.

The other small clusters that co-cited closely with Field A include the chromatin-remodelling protein Brg1 (#4), von (#11), p21 (#13) and human telomerase (#15 and #16). As a ribosomal nuclear protease, new tumour marker and anticancer target, telomerase has become a hot topic in breast cancer research. Telomerase has the ability to maintain the normal karyotype after a cell exceeds its normal lifespan in vitro, revealing the mechanism of abnormal division in cancer cells (30).

Research front 2

Many studies have indicated that angiogenesis is the basis for tumour metastasis, and malignant tumours often grow slowly without angiogenesis.

Our map shows that many scholars are interested in metastatic renal cell carcinoma and metastatic colorectal cancer. With the continuous exploration of the mechanism of renal cell carcinoma development, many of the hypoxia-inducible factor (HIF)-associated proteins, including VEGF (vascular endothelial growth factor), PDGF (platelet-derived growth factor) and TGF-α (transforming growth factor-α), are involved in renal cell carcinoma. Endothelial cell receptors, periderm cell receptors, and tumour cell receptors have also been shown to play roles in renal cell carcinoma.

In recent years, the molecular-targeted therapy of metastatic renal cell cancer and metastatic colorectal cancer has become a hotspot in oncology research. For example, bevacizumab, a main molecular-targeted drug, is a monoclonal antibody against vascular endothelial growth factor (31). In addition, as an inhibitor of angiogenesis, cilengitide has a function in preventing the production of blood vessel production (32).

Presently, molecular-targeted drug treatment is a new approach to inhibit tumour development, invasion and metastasis. There are many outstanding research results in this research area. For instance, the investigations of Paez et al(33) and Lynch et al(34) indicate that EGFR mutations can predict sensitivity to gefitinib. These studies provide the foundation for the application of this molecular-targeted drug.

In Fig. 3, two clusters can be seen, endostatin (#44) and VEGF (#45), each of which have an intense co-citation relationship with metastatic colorectal cancer and metastatic renal cell carcinoma, indicating that these clusters are closely related. In 1971, Folkman (35) first proposed that tumour growth and metastasis rely on angiogenesis. According to recent relevant publications, the inhibition of tumour angiogenesis and induction of tumour hibernation or apoptosis are effective methods for treating tumours. For example, endostatin plays an anti-angiogenic role and exhibits broad-spectrum, low toxicity and non-resistant characteristics. The application of endostatin is a new strategy for treating cancer, and anti-VEGF therapy is predicted to become a mainstream therapy for cancer treatment.

Fig. 3 also shows that tumour metastasis co-cites with another cluster (protease ADAM9), indicating that these clusters are closely related. The expression of ADAM9 in colon cancer tissue was found to be significantly higher than that in adjacent normal tissue and has a correlation with the disease progression of metastatic colon cancer and tumour angiogenesis (36).

Research front 3

Apoptosis, also known as programmed cell death, was first described in 1972 by Kerr (37), an American pathologist. In recent years, researchers have been paying close attention to apoptotic mechanisms. Commonly, the occurrence of tumours results from the unlimited growth of tumour tissue, which is caused by unregulated apoptosis. Mitochondria, the only organelles containing their own genetic material in mammalian cells (38), play a very important role in the process of tumour cell apoptosis. Caspase activation is closely linked with mitochondrial outer membrane permeabilisation in the mitochondrial pathway of apoptosis (39).

Based on in-depth studies of apoptosis and tumour pathogenesis, a number of ligands and receptors that induce tumour cell apoptosis have been identified, all of which belong to the tumour necrosis factor (TNF) supergene family. Regarding the tumour necrosis factor family, the TRAIL (TNF-related apoptosis inducing ligand) system is one of the research hotspots of recent years. The overexpression of Trail (40) can selectively kill a variety of tumour cells yet shows no significant toxicity to most normal cells, suggesting its possible use as a new anticancer drug. A study published by Walczak et al(41) demonstrated that the growth of a human mammary adenocarcinoma cell line could be inhibited by LZ-TRAIL without significant toxicity to normal tissues. Another outstanding experiment performed by Ashkenazi et al(42) showed that Apo2L (or TRAIL) has potent anticancer activity but leaves normal cells unharmed.

Research front 4

With the development of tumour immunology and molecular biology, the research areas involving the interaction between the tumour and the body, tumour immune tolerance and tumour antigen identification have made great progress and have contributed to the development of cancer vaccines. To date, many tumour vaccines have been developed via animal experiments and clinical trials. Various effective vaccines induce cytotoxic T-lymphocytes with adverse reactions. Therefore, as an efficient tumour immunotherapy with low toxicity, tumour vaccines are potentially effective drugs that must be developed.

A study by Cao et al(43) suggeted that it is possible to fight solid tumours by modulating antitumour immunity. Their research provided a solid foundation for theories on tumour vaccines. An example of a cancer vaccine is the immunodominant peptide of the gp100 melanoma-associated antigen constructed by Rosenberg et al(44). His experiment initiated the development of novel cancer immunotherapies. In another example of an effective tumour vaccine, BERH-2 tumours can be eradicated by hybrid cells made from the fusion of BERH-2 rat hepatocellular carcinoma cells with activated B cells (45).

As mentioned above, the difference between tumour vaccines and conventional vaccines is that tumour vaccines are active immunotherapies for treatment and do not prevent infection with pathogens. With the development of tumour immunology, the tumour mechanism of immune tolerance has been gradually recognised, and an increasing number of tumour antigens have been identified. These antigens are helpful in the research and development of tumour vaccines. At present, the majority of tumour vaccines, such as the gastric tumour cell vaccine, lung tumour cell vaccine, and melanoma cell vaccine, have been tested in animal models, and only a few have been studied in clinical trials, such as the cervical cancer vaccine and breast cancer vaccine. Preparation of tumour vaccines has gradually improved and developed to include genetic modifications and cell fusion methods. In the near future, tumour vaccines may become a new means of cancer treatment after surgery, radiotherapy and chemotherapy.

In conclusion, having analysed the results, we understand that 30 major international oncology journals do not represent the overall productivity in the field of oncology; thus, there may be a few disadvantages regarding the research front based on the authors’ outputs. Otherwise, compared to the disadvantages, this network offers many advantages as it is more practical, easy to handle, and effective, and will be more instructive for scientific research evaluation work.

However, by mapping the knowledge domains of oncology research, we found that the citation frequency for the US accounted for 53.8% of the total citation frequency of the top 20 core countries performing oncology research. China is the sole developing country to enter the top 10. Within the core institutions of oncology research, the US accounted for 18 of the top 20 institutions, and its total citation frequency accounted for 93.5% of the top 20. The research strength of the University of Texas, the NCI and Harvard University has steadily grown for nearly 10 years, and most countries and institutions have separate and distinct research directions. It is clear that the United States is the leading country in oncology research.

We found that the current oncology research fronts are focused in four fields: i) the mechanism of abnormal oncogene expression; ii) tumour metastasis and angiogenesis; iii) the relationship between cancer cells and apoptosis; and iv) tumour vaccines. The four research frontiers included 55 branch fields. We also identified the 36 most collaborative academic communities and determined that their oncology research strength is basically in balance. Multiple myeloma, angiogenesis, and acute lymphocytic leukaemia were the primary focuses of research collaborations in oncology from 2001 to 2010.

After our analysis of the main research centres with respect to institutions and countries, we traced oncology articles published in the studied time frame. By document co-citation analysis, we accurately identified the research fronts, hotspots, and classic milestone publications that provide the foundation in the field. From our co-authorship analysis, we determined the most collaborative academic communities. Through the interpretation of our knowledge maps, we depicted the overall image and trends of oncology research progress in an objective, scientific and systematic manner. We also demonstrated the inherent mainstream areas and knowledge structure in the oncology discipline so that countries with developing research programs can track the international research forefronts, grasp the correct direction for their research, and identify entry points for their own studies. These maps also provide scientific evidence and suggestions for policymakers to select the most prolific academic groups and individuals in the oncology research community and establish a more efficient system for managing and financing oncology research in the future. The study not only provides support for important decisions in the science and technology field but also supplies a basis for technology planning and evaluation and a quantitative basis for the selection of research projects. An improvement in international oncology research has important theoretical significance and academic value.

There are many issues worthy of further analysis in oncology research. We are planning to supplement our current data with that from 1981 to 2000 to further study the changing trends in the oncology research fronts and hotspots and the changing scientific laws in the field of oncology. Because the present scientometric methods have not identified relevant disease-specific or clinical activities, we will combine the meta-analysis with the present methods to evaluate disease identification in our next study.

Acknowledgements

This study was supported by the National Nature Science Foundation of China under grant numbers, No. 71240006 and No. 71103114; the Taiyuan City Science and Technology Innovation Project Program under grant number, No. 110271. In addition, the authors would like to thank other member of the authors’ team for helpful discussions, particularly Professor Dawei Guo.

References