Co-occurrence theory holds that if two items appear together in the same intentional unit (such as author, keyword, institution, English), this indicates a strong correlation between the two projects, such as similar semantic connotations, interaction between the items, or cooperation [47,48]. Similarly, if multiple disciplines appear together in the subject category field of bibliographic data, we can speculate that there is cross-cooperation between these disciplines [45,49]. By extracting the cross-cooperative relationships of all subject categories, a complete cooperative network is formed. The following analysis of the structure can reveal hidden cooperation features and laws [40,50].

An important part of co-occurrence analysis is to analyze the network structure formed by the co-occurrence relationship for the overall and individual network indicators. We introduced the bibliographic data into the Science of Science Tool，version 1.2 beta (Cyberinfrastructure for Network Science Center, Indiana University, Bloomington, Indiana, United States), to extract the subject category field, count the number of subjects, and calculate the co-occurrence frequency between any two subjects [51]. This means that two subjects appear together in the same bibliographic data. For co-occurrence, the total frequency of co-occurrence is equal to the amount of bibliographic data containing the two subjects. On the basis of extracting the discipline and its co-occurrence relationship, a cross-disciplinary cooperation network was generated and exported as a “.net” file. In the co-occurrence network file, the points and edges represent the disciplines and their cooperative relationship, respectively, and the frequency of appearance and the frequency of co-occurrence are weighted.

In general, the network analysis focuses on its largest connected subgraph because the isolated or unconnected points do not reflect the main connotation. We used SCI2 to eliminate the isolated points (disciplines) in the disciplinary cooperation network to generate the largest connected subgraph. The new net file was used as the basis for subsequent analysis. The maximum connected subpicture file was imported into Pajek [52] for network index calculation (including centrality, density, and aggregation coefficient), and a topology map of the cooperative network was generated. Network indicators (including integrated network indicators and individual network indicators) are the embodiment of the cooperative network structure, reflecting the position and function of the discipline in the cooperative network as well as the laws and trends of cross-disciplinary cooperation. We can even speculate on the laws and trends of the cooperative discipline network.

It is worth noting that the nodes in the co-occurrence network exhibit certain aggregation characteristics due to the different connection distributions. The nodes that are grouped into the same class form a community, indicating that the nodes are similar in a certain aspect. In the same way, if the disciplines in the interdisciplinary cooperation network are divided into the same class due to the cooperative relationship, this indicates that the intensity of crossing cooperation between them is strong. We can also infer that the disciplines mentioned above have unity in their research direction and theme. In this study, we used the Louvain community partitioning algorithm in Pajek [53] to divide the disciplinary cooperation network into numerous different communities and explore the characteristics of precision medicine research in terms of disciplinary cooperation.

Subject category and its co-occurrence relationship in Web of Science bibliographic data provide strong support to describe the extent of interdisciplinary cooperation or interdisciplinarity [45,49]. We also used String’s diversity index and the specialization index to calculate the interdisciplinary degree of precision medicine research [46,54].

String’s diversity index calculates the diversity of discipline cooperation. The greater the value is, the greater the interdisciplinary degree is. The calculation formula is as follows:

and is the proportion of the occurrence frequency of subject and to the sum, and is the degree of difference between subject and ; its value is calculated by Formula 2. Furthermore, is Salton's cosine similarity between the two disciplines [55]; its value is calculated by Formula 3. Formula 3 calculates the similarity of subjects based on the number of co-occurrences between one subject and the other disciplines as well as the similarity between two associated disciplines. It can transform the disciplinary cooperation network into a co-occurrence matrix and into a cosine similarity matrix, indicating the similarity of any two disciplines.

The specialization index is used to describe the concentration level of disciplinary cooperation; its meaning is opposite to String’s diversity index. The larger the specialization index is, the fewer disciplines involved in the cooperation. This indicates that overall cooperation is limited. The formula of the specialization index is as follows:

where is the frequency of occurrence of subject .

Combining the above two indicators, the discipline cooperation status of precision medicine research can quantitatively reveal the degree of yearly cooperation of one discipline. Then, we can discover the chronological change of the disciplinary cooperation of precision medicine research.

Visualization has the advantage of displaying the co-occurrence network structure and posture, thus helping us better understand the meaning of the research object. Due to the superiority of VOSviewer [56] in terms of visualization effects, we selected it to show the subject cooperation network, including the overall network at the community level and the network of each single community. In addition, we revealed the chronological changes of subject cooperation. In this study, we divided the bibliographic data according to age and introduced Cortext to generate interconnected strip diagrams to show the chronological characteristics of interdisciplinary cooperation. Because of the various cooperation intensities and distribution structures, there were significant differences in the subject cooperation community. The diversity between the communities is reflected in the two indicators, such as density and average centrality. It can quantitatively display the relative position and development status of the cooperative community in the whole subject cooperation network. Based on the above two indicators and the sum of disciplinary frequency in the community, we drew a strategic diagram to intuitively show the relative development trend of the discipline cooperation community [57]. The strategy map uses the average of all community densities and centralities as the origin, with the centrality as the x-axis and the density as the y-axis, dividing the map into 4 quadrants to show the differences between the communities. The centrality reflects the degree of association between a community and other communities. The higher the value, the more central the community is in the entire network. The density reflects the closeness between the communities. The higher the value, the closer the internal association is and the more mature the research field is. Each community is distributed in 4 quadrants due to its centrality and density. The community in the first quadrant, with a high degree of centrality and density, is the core of the whole research and the most mature development; the community in the second quadrant, with a lower center and higher density, is not the core but is mature in the whole research. The community in the third quadrant has low centrality and density; it is neither the core of the whole research nor immaturely developed. The community in the fourth quadrant has higher centrality but lower density; it is the core of the whole study, but the development of the community is not mature. The different distribution of discipline cooperation communities in the quadrant represents their relative development status.

The interdisciplinary network of precision medicine research is the largest connected subgraph, which contains 433 edges, representing interdisciplinary cooperation. It is worth noting that the intensity of interdisciplinary cooperation (co-occurrence frequency) varies from 1 to 104. Cooperation, which has a co-occurrence frequency greater than or equal to 30, accounts for 51.4% of the total (2292/4458) and mainly focuses on medical informatics, cell biology, health care, sciences and services, biochemistry and molecular biology, and genetics and heredity (as shown in Table 2). This means that the interdisciplinary cooperation mentioned above is the mainstream of current precision medicine research.

According to the overall network indicators of interdisciplinary cooperation (Table 3), the density value indicates that the discipline cooperation of current precision medicine research is poor; the cross-cutting nature of precision medicine research is still immature. At the same time, the degree centralization and closeness centralization of the network are not high, showing that the concentration of discipline cooperation in precision medicine research is not high and is scattered. In other words, the influence or dominance of a discipline on the whole cooperation network is not obvious. Network betweenness centralization is high, indicating that most interdisciplinary cooperation is likely to be indirect; that is, interdisciplinary cooperation requires other disciplines as a “bridge”. This makes the distance between two disciplines in the cooperative network long and makes the discipline cooperation network loose. The above results are also reflected in the clustering coefficient, the value of which is higher than the overall degree centralization. This means that the disciplines are gathered into different clusters due to the different cooperation structures, and interdisciplinary cooperation within the clusters is above the overall level. Therefore, we can speculate that in some subject directions, precision medicine has formed a relatively stable multidisciplinary cooperation, and researchers in various disciplines have reached a basic consensus on certain directions.

In the disciplinary cooperation network, the network indicators (degree centralization, closeness centralization, and betweenness centralization) of each discipline reflect its position and role in the entire network. As shown in Table 4, pharmacology and pharmacy, oncology, genetics and heredity, biochemistry and molecular biology, neurosciences and neurology, engineering, research and experimental medicine, biotechnology and applied microbiology, and cell biology have higher network indexes, indicating that these disciplines are at the core of the network. They are the most cooperative, their cooperative patterns are direct, and their cooperation is in short paths. It can be suggested that these disciplines play a leading role in current precision medicine research, and cooperation among these disciplines is the mainstream of current precision medicine research. In contrast, except for the higher betweenness centralization of pharmacology and pharmacy, the betweenness centralization of other disciplines is low. This indicates that pharmacology and pharmacy has played an important “bridging” role in the interdisciplinary cooperation of precision medicine research.

The disciplinary cooperation network of precision medicine research is well divided into 7 communities, with a module degree of 0.4343. This indicates the strong preference for disciplinary cooperation in precision medicine research. We discover some research directions of precision medicine with close cooperation of multiple disciplines. There is a significant difference between these directions represented by communities. As shown in Table 5, the disciplinary cooperation network for precision medicine research includes the following 5 communities:C1-oncology community, including cardiovascular system and cardiology; cell biology; respiratory system; radiology, nuclear medicine, and medical imaging; hematology; gastroenterology and hepatology, etc; C2-pharmacology and pharmacy community, including neurosciences and neurology; psychiatry; endocrinology and metabolism; toxicology; psychology; integrative and complementary medicine, etc; C3-health care sciences and services community, including mathematical and computational biology; computer science; medical informatics; engineering; public, environmental, and occupational Health, etc; C4-genetics and heredity community, including biochemistry and molecular biology; biotechnology and applied microbiology, etc; C5-biomedical social sciences; C6-research and experimental medicine; C7-immunology communities.

From the perspective of the scale of cooperation, the current disciplinary cooperation of precision medicine research is clearly divided into three levels: the largest is the C1 direction, followed by the C2, C3, and C4 directions, while the C5, C6, and C7 directions are smaller. In other words, in the past decade, precision medicine research has focused on oncology, pharmacology and pharmacy, health care sciences and services, and genetics and heredity, which represent the mainstream direction of research. However, biomedical social sciences, research and experimental medicine, and immunology research in the discipline are still weak.

The interdisciplinary structure of precision medicine research needs to be presented through a visual map, as shown in Figures 4 and 5. Figure 4 presents a network of disciplinary cooperative communities. Each point represents a community and is distinguished by different colors. The size of the points represents the sum of the frequencies of all disciplines in the community (the number in brackets in the figure), which can be regarded as the scale of the discipline cooperation. The larger the node is, the larger the scale of cooperation is. Each edge represents the cooperative relationship between the communities. The thickness of the edge represents the intensity of cooperation, which is proportional to the sum of the number of co-occurrences between the two communities. Moreover, there is extensive cooperation between the C1-oncology community, the C3-health care sciences and service community, and the C6-research and experimental medicine community, as well as cooperation between the C1-oncology community and the C2-pharmacology and pharmacy community. However, C5-biomedical social sciences and C7-immunology are relatively isolated and less collaborative with other communities.

Figure 5 shows the characteristics of cooperation among the internal disciplines of each community. In Figure 5, different colors indicate their own community; each node represents a discipline, and its size is proportional to the frequency of the discipline. Each edge represents the relationship between disciplines, and the thickness of the edge represents the number of co-occurrences between disciplines. This indicates that precision medicine research is inclined to concentrate on disciplinary cooperation. The network indicators of the disciplinary cooperation communities shown in Table 6 also support our conclusions presented above.

The density of each community exceeds the overall network. While the C5-biomedical social sciences, C6-research and experimental medicine, and C7-immunology communities are too small to be comparable, other communities have higher densities. In particular, the C4-genetics and heredity community has a relatively high degree of centrality and density, indicating that in precision medicine studies, the disciplines involved in this community and their related research directions are the core of precision medicine research, and they are widely related to other disciplines and research directions. Moreover, the communities of C1-oncology, C2-pharmacology and pharmacy, and C3-health care sciences and services are the main and core components of current precision medicine research disciplines and have a greater impact on the entire study.

Although the interdisciplinary cooperation mode changes every year, interdisciplinary cooperation continues. Considering the different scales and levels of interdisciplinary cooperation in different stages, we attempt to reveal the evolution of interdisciplinary cooperation in precision medicine research in two stages: 2010-2014 and 2015-2019 (as shown in Figure 8 and Figure 9). To show the time continuity, the second stage of the evolutionary graph starts in 2014. According to the centrality and density of the interdisciplinary cooperation community, we mapped the interdisciplinary cooperation community to the 2-dimensional strategic map with the average value of the centrality and density of all communities as the quadrant origin to reveal the relatively low position and development trend of the interdisciplinary community in precision medicine research (Figure 10).

In the first stage (2010-2014), the interdisciplinary cooperation of precision medicine research is relatively stable. There are 3 evolutionary streams: (1) mathematical and computational biology and medical information, including engineering, computer science, and mathematics; (2) integrated and comprehensive medicine and plant sciences, including neurosciences and neurology, pharmacology and pharmacy, psychiatry, psychology, nutrition, dietetics, food science and technology; and (3) research and experimental medicine and cell biology, including materials science, chemistry, biochemistry and molecular biology, cell biology, pediatrics, physiology, genetics and condition, toxicology, medical history, history and philosophy of science, and oncology.

The continuity of each stream of cooperation is good, but the scales are different. It can also be observed that the interdisciplinary cooperation of precision medicine research at this stage focuses on the two main streams: (1) integrative and complementary medicine and plant sciences and (2) research and experimental medicine and cell biology. It is worth noting that research and experimental medicine and cell biology have been fused and differentiated many times. For example, in 2013, pediatrics and hematology, research and experimental medicine, and genetics and heredity were fused into another stream, research and experimental medicine and cell biology. Interestingly, in 2014, the streams of research and experimental medicine and cell biology were also fused into the completely new streams of medical laboratory technology, anatomy and morphology and oncology and pediatrics. The characteristics of the trends in disciplinary evolution show the instability and expansion of interdisciplinary cooperation in precision medicine research.

In the second phase (2015-2019), according to the interdisciplinary cooperative community division in 2014 (Figures 8 and 9), the interdisciplinary cooperation at this stage continued the status of the first phase and continued to evolve into 2019. At this stage, there are three major evolutionary streams. First, medical informatics and mathematical and computational biology already has a relatively stable relationship with computer science and other disciplines. Second, history and philosophy of science and medical ethics, including social sciences, science and technology, chemistry, physics, health care sciences and services, medical ethics, social issues, and biomedical social sciences, is integrated with imaging science and photographic technology, optics, and physics. Finally, plant sciences and integrative and complementary medicine and the context of oncology and pediatrics is a continuation and fusion of the previous phase, including research and experimental medicine, biotechnology and applied microbiology, neurosciences and neurology, pharmacology and pharmacy, cardiovascular system and cardiology, transplantation, psychology, and surgery.

At this stage, interdisciplinary cooperation focused on the fused stream of plant sciences and integrative and complementary medicine and oncology and pediatrics. It is worth noting that pharmacology and pharmacy and neurosciences and neurology (2017) and surgery and pediatrics (2018) became the two major communities in the stream mentioned above. This indicates a shift in the direction of interdisciplinary collaboration in precision medicine research.

At the same time, there are a few isolated communities or intermittent evolutionary streams, such as immunology and microbiology, physics and spectroscopy, food science and technology, and government and law. These streams or communities do not effectively continue. These changes are worth considering in precision medicine interdisciplinary collaboration.

According to the indicators of community centrality and density, the developmental trend of the interdisciplinary cooperative community of precision medicine research is shown in Figure 10. C3 and C4 are in the first quadrant. Due to their high centrality and density, we can speculate that these communities are the core of interdisciplinary cooperation in precision medicine research, and their cooperative status is relatively stable and mature. C5, C6, and C7 are in the second quadrant. Although their cooperative state is relatively stable and mature, the disciplines involved are no longer the core of current precision medicine research. It is worth noting that the location of the C6 community is very close to the original point and has great potential for development. It is likely to be the core community in the future. C1 and C2 are in the third quadrant as the two largest cooperative communities. Their cooperative state is unstable and mature, and it is not the core discipline or direction of the entire precision medicine research.