Global research Trends in Tyrosine Kinase Inhibitors: A Co-Word and Visualized Study

Background: Tyrosinase inhibitors (TKIs) have achieved revolutionary results in the treatment of a wide range of tumors, which have brought out a lot of literature in this field every year. And some reviews provide a great value for us to understand TKIs. However, there is a lack of studies on the knowledge structure and bibliometric analysis in TKIs research. Objective: This paper aims to investigate the knowledge structure, hotspots, and trends of evolution of the field of TKIs by coword analysis and literature visualization. Methods: We obtained all academic papers about TKIs published in 2016-2020 from the Web of Science. Based on counting keywords from those papers, we generated the co-word networks by extracting the co-occurrence relationships between keywords, and segmented communities to identify the sub-directions of TKIs research by calculating the network metrics of the overall and local networks. We also mapped the association network topology, including the network within and between TKIs sub-directions, to reveal the association and structure among varied sub-directions. Finally, evolution venation and strategic diagram were generated to reveal the trends of TKIs research. Results: We obtained 6,782 unique words (total frequency 26,175) from 5,584 paper captions. Finally, 296 high-frequency words were selected with a threshold of 10 after discussion, the total frequency of which accounted for 65.41%. The analysis of burst disciplines revealed a variable number of burst words of TKIs research every year, especially in 2019 and 2020, such as HER2, Pyrotinib, Next-generation Sequencing, Immunotherapy, ALK-TKI, ALK Rearrangement, etc. By network calculation, the TKIs co-word network was divided into six communities: C1-Non-small Cell Lung Cancer, C2-Targeted Therapy, C3-Chronic Myeloid Leukemia, C4-HER2, C5-Pharmacokinetics, and C6-ALK. The venation diagram revealed several clear and continuous evolution trends, such as Non-small Cell Lung Cancer venation, Chronic Myeloid Leukemia venation, Renal Cell Carcinoma venation, Chronic Lymphocytic Leukemia venation, etc. In the strategic diagram, C1-Non-small Cell Lung Cancer was the core direction located in the first quadrant, C2-Targeted Therapy was exactly at the junction of the first and fourth quadrants that meant C2 was developing, and C3-Chronic Myeloid Leukemia, C4-HER2, and C5-Pharmacokinetics were all immature that all located in the third quadrant. Conclusions: Using co-word analysis and literature visualization, we revealed the hotspots, knowledge structure, and trends of evolution of TKIs research during 2016-2020. TKIs research mainly focused on targeted therapies against varied tumors, particularly against NSCLC. The attention on CML and pharmacokinetics was gradually decreasing, but the heat of HER2 and ALK was rapidly increasing. TKIs research had shown a clear development path: TKIs research was disease-focused and revolved around "gene targets/targeted drugs/resistance mechanisms". Our outcomes will provide sound and effective support to researchers, funders, policymakers, and clinicians. (JMIR Preprints 29/10/2021:34548) DOI: https://doi.org/10.2196/preprints.34548 https://preprints.jmir.org/preprint/34548 [unpublished, non-peer-reviewed preprint] JMIR Preprints Hu et al


Preprint Settings
1) Would you like to publish your submitted manuscript as preprint?Please make my preprint PDF available to anyone at any time (recommended).
Please make my preprint PDF available only to logged-in users; I understand that my title and abstract will remain visible to all users.Only make the preprint title and abstract visible.
No, I do not wish to publish my submitted manuscript as a preprint.2) If accepted for publication in a JMIR journal, would you like the PDF to be visible to the public?
Yes, please make my accepted manuscript PDF available to anyone at any time (Recommended).
Yes, but please make my accepted manuscript PDF available only to logged-in users; I understand that the title and abstract will remain v Yes, but only make the title and abstract visible (see Important note, above).I understand that if I later pay to participate in <a href="http

Introduction
Tyrosine kinases (TKs) are a collective term for dozens of kinases encoded by multiple genes, which can phosphorylate tyrosine residues in cells 1 .Based on varied cellular localizations, the TKs family is divided into receptor tyrosine kinases (RTKs) 2 and non-receptor tyrosine kinases (non-RTKs) 3 .RTKs consist of 20 subfamilies (e.g., epidermal growth factor receptor (EGFR) 4 , vascular endothelial growth factor receptor (VEGFR) 5 ), whereas non-RTKs include 10 subfamilies such as ABL, SRC, and CSK 6 .TKs have the common activity to catalyze the transfer of γ-phosphate groups on adenosine triphosphate (ATP) to the tyrosine residues of a variety of target proteins 1,[3][4][5][6][7] , and this process plays a key role in signal transduction in the cell.Abnormal activities of TKs are closely associated with proliferation, invasion, metastasis, apoptosis, and tumor angiogenesis in non-small cell lung cancer (NSCLC) 8 , chronic myeloid leukemia (CML) 2,9 , and many other tumors.Therefore, TKs have become excellent targets for tumor therapy.
Tyrosine kinase inhibitors (TKIs) are referred to as a class of small-molecule compounds that can specifically inhibit TKs.They can penetrate through the cell membrane and block the signaling pathway of tumor proliferation, and some TKIs can also inhibit angiogenesis 1,10 .TKIs have revolutionized the treatment of a variety of tumors [10][11][12] , for example, imatinib has been a typical pioneer in successfully translating oncogene research into molecular targeted therapy.Now, TKIs have developed to the fourth generation, which aims to overcome drug resistance due to mutations of T790M and C797S) 13 .More than 30 small-molecule TKIs have been approved for marketing by the U.S. Food and Drug Administration (FDA), and hundreds of drug candidates are in various stages of clinical trials [13][14][15] .Therefore, this article aims to understand the development process of TKIs research, identify the main research directions, and analyze the potential research hotspots.
Co-word analysis is a content analysis method to study the knowledge structure and evolutionary patterns of various fields.It can facilitate researchers to identify hotspots, composition, paradigms, and evolution of a field by calculating the word pairs and co-occurrence of noun phrases in the literature [16][17][18] .This method has been used widely in medical bibliometric analysis, such as precision medicine 16 , neonatal ischemic-hypoxic encephalopathy 19 , stem cell research 20 , neural stem cells 17 , tumor immunotherapy 18 , disaster medicine 21 , medical big data 22 , surgical robotics 23 , epilepsy genetics 24 , etc.We propose to use the co-word analysis to explore the knowledge structure, evolution trends, and associations among sub-topics of TKIs research, aiming to help clinicians and scholars have a better understanding of TKIs and to give suggestions for research and usage of TKs.

Literature review 2.1 Background
In recent years, targeted therapies have become a hotspot in the development of anti-tumor drugs with their advantages of high selectivity and low side effects 25,26 .TKIs are revolutionary targeted drugs that inhibit tumor proliferation by interfering with or inhibiting specific proteins within cancer cells, thus exerting prominent anti-tumor effects [1][2][3] .Among them, imatinib was the first targeted antitumor drug 27 , which was first approved in 2001 for the treatment of BCR-ABL-positive and Philadelphia chromosome-positive CML 3 , 6 .And then, the first, second, and thirdgeneration TKIs, represented by gefitinib, dasatinib, and osimertinib, have been validated in hundreds of clinical trials and approved for marketing 1 0 -1 2 , 2 8 , 2 9 .
Genetic testing has been developed rapidly.Next-generation sequencing (NGS) allows for the sequencing genome and exome within days and makes it possible to identify patients with druggable mutations quickly and precisely 30 .Meanwhile, multidisciplinary collaboration between pharmacology and clinical science has brought a leap forward in basic research and clinical applications of TKIs.First, tumor-targeted therapies are the most established area for TKIs, especially in the treatment of lung cancer 12,15,31,32 , and leukemia 10,11,28,29 .It's demonstrated that TKIs improved the quality of life and extend survival in patients with advanced NSCLC 33 .Imatinib and gefitinib have become first-line drugs due to their outstanding clinical efficacy in patients with BCR-ABL-positive CML 10,11,28,29 .Second, clinical trials of various drugs targeting HER2 and ALK (e.g., trastuzumab 34,35 , patolizumab 36 , ceritinib 37 ) have manifested excellent effects.Third, pharmacokinetics is another focus of TKIs research.Optimization and selectivity study is an important direction for continuing clinical trials after the launch of many TKIs.And individualized blood concentration monitoring is also important for patients with poor efficacy or severe side effects 38 .

2.2Previous efforts
In recent years, TKIs have been widely used for tumor-targeted therapies.Numerous research efforts helped clinicians and scholars better understand TKIs and facilitated the clinical translation of study outcomes.
Based on recent reviews, the current status of TKIs research is summarized as follows: First, TKIs resistance is becoming increasingly prominent, of which genetic mutations (e.g., T790M 39,40 , C797S 13 , D761Y 41 , L747S 42 ) are the main cause.It has become a key direction to find new molecular mechanisms of TKIs resistance and to establish individualized dosing regimens.Second, the application of drugs such as Erlotinib 43 , Osimertinib 15,33 , and Gefitinib 44 has gradually matured and occupied an important position in the treatment of various tumors such as NSCLC 11,12 .Third, drugs targeting HER2 and ALK continue to emerge, which offers new hope for solving the plague of drug resistance 45,46 .To date, hundreds of new TKIs candidates are in various stages of clinical research 47 .

The rationale for the study
Research on TKIs continues to grow to benefit more patients.However, there is still a lot of uncharted territories to explore in TKIs research.How to discover new biomarkers of TKIs?How to select TKIs with better clinical effects and fewer side effects for targeted therapy?How to overcome multidrug resistance in patients with tumors?All these questions need scientific bibliometric analysis based on the results of TKIs research.The purpose of our study is to address the following questions: (1) What is the overall knowledge structure of TKIs research?
(2) What are the sub-directions of TKIs research and how do they interact with each other?
(3) What are the evolutionary status and development trends of TKIs research in the temporal dimension?

3.1Data collection and processing
It is well known that Web of Science Core Collection (WOSCC) is the most extensive and comprehensive academic literature database, so we used subjects (including "Tyrosine kinase inhibitor, Tyrosine kinase inhibitors, TKI, TKIs, Tyrosine kinases inhibitors, Tyrosine kinases inhibitors") in WOSCC to precisely search all literature about TKIs by limiting the period to 2016-2020 and the literature types to journal papers, reviews and conference papers.The specific search formula is "(TS= ("Tyrosine kinase inhibitor" OR "Tyrosine kinase inhibitors" OR "TKI" OR "TKIs" OR "Tyrosine kinases inhibitors" OR "Tyrosine kinases inhibitor")) AND LANGUAGE: (English) Refined by: DOCUMENT TYPES: ( ARTICLE OR REVIEW OR PROCEEDINGS PAPER ) Timespan: 2016-2020.Indexes: SCI-EXPANDED, SSCI, A&HCI, CPCI-S, CPCI-SSH, ESCI, CCR-EXPANDED, IC".
A total of 13,895 documents were retrieved and exported in Tab-delimited (Win) format.Next, the records containing search terms in the titles or keywords were retained, while the records without keywords and the records with search terms appearing only in the abstracts were excluded 48,49 .Finally, 5,584 records were obtained for the subsequent co-word network analysis.(Figure 1) Since there are irregularities and inconsistencies in the writing of keywords submitted by the authors themselves in WOS, it is necessary to pre-process them.First, this paper aims to depict the research status of TKIs by using other terms associated with TKIs, so TKI itself, the synonyms, and hypernyms of TKI were removed.Keywords whose meaning is broad (e.g., review, development, problem, etc.) were also removed.Secondly, the co-word analysis generally targets high-frequency keywords and their relationships, while keywords with very low frequencies cannot reflect the main direction of the current research.Therefore, this paper firstly generated a list of keywords by frequency descending order and then defined the threshold of high-frequency keywords according to the cumulative percentage of frequencies in the list 50 .Third, keywords with frequencies below the threshold were merged into the words with the closest meanings, and words with the same meaning and different forms were also merged, such as "BCR-ABL TKI" to "BCR-ABL" and "Epidermal Growth Factor Receptor" to "EGFR".Fourth, after de-duplicating the merged keywords, a new list of keyword frequencies was generated.

Network construction and analysis
The keywords in a paper are an accurate description of its main content, so mining keywords and their relationships can help reveal the hidden connotation of a research field 51 .If two words cooccur in the same connotation unit (e.g., keywords in a paper), they are related or similar in connotation and have consistency in connotation expression.Their co-occurrence frequency is equal to the number of papers that contain them at the same time, and the greater the frequency, the stronger the semantic association between them 52 .By constructing co-word networks and performing structural analysis and visualization, co-word analysis can effectively reveal the underlying connotations, research structures, and even evolutionary trends of a research field 53 .
In this paper, the above pre-processed data were imported into SCI2 54 for frequency statistics and co-word network generation (.net format).Then, the .netfile was imported into the network analysis tool Pajek 55 to calculate network indicators, including centralization (centrality) 56 , density 57 , and the clustering coefficient 58 , and to perform community segmentation to identify major subdirections.The centralization refers to the centripetal or consistency of the co-word network as a whole, while centrality reflects the keywords' position in the network and its ability to influence and control the network 56 ; The density represents the degree of association of the network as a whole, and the stronger the association, the more mature the research field; the clustering coefficient reflects the possibility that words will cluster into classes depending on the association and its strength, and the possibility that the network will be distinctly divided into several sub-networks or sub-classes.Combined with the community segmentation algorithm (Louvain) 59 , the co-word network will be divided into distinctive communities, each of which represents a sub-direction.with strong ties within the communities and loose ties between the communities, reflecting a greater concentration or consistency in the connotative associations of words within the community.Keywords are tightly linked within communities and loosely linked between communities, reflecting that keywords possess more focused or consistent connotations within communities.

Mapping and Visualization
To show the structure and characteristics of the TKIs research more intuitively and clearly, we visualized the topology, evolutionary venations, and development trend of the co-word network.
First, the visualization of the network topology.In this paper, Vosviewer 60 was used for the multi-level presentation of co-word networks and communities, including the inter-community and intra-community association network graphs.In the network graphs, nodes represent keywords or communities, and edges represent co-occurrence relationships between words or communities.The size of nodes and the thickness of lines are proportional to the frequency of keywords and the scale of communities, respectively, and the nodes and lines belonging to different communities are distinguished by different colors.These network diagrams visualize the importance and association relationships of keywords or communities in TKIs and help to analyze the distribution and structural characteristics of TKIs research.
Second, the visualization of evolutionary venations.In this paper, we divided each year's records into several communities.Then we used Cortext 61 to calculate the overlapping relationships between communities in adjacent years and connected them through "tubes".In the tube diagram, bars of different colors and sizes represent communities of different sizes, and the tubes connected by several bars represent the continuation of the research theme, which can be considered as evolutionary venations.The evolutionary trends of TKIs research over time are visualized by graphically characterizing the continuity, convergence, and divergence of communities.
Third, the visualization of the developmentary degree of the sub-directions of TKIs research.These research communities can be considered as sub-directions of TKIs research, and each community or sub direction exhibits specific development status depending on the density and centrality.So, we drew a two-dimensional strategic diagram based on the calculation of the density and centrality of each community.The strategic diagram took centrality, which represented the core degree of research directions in TKIs, as the horizontal axis, and density, which represented the developmental maturity of research directions, as the vertical axis, and the mean of community density and centrality as the origin.Ultimately, communities were mapped into 4 quadrants to visualize the degree of centrality and maturity of different research directions in TKIs.
Fourth, the visualization of burst words.The changes in keyword frequency fluctuate significantly, with some of the words appearing in sudden bursts, reflecting the existence of distinct epochal characteristics of TKIs research.Therefore, we detected keyword bursts and combined their burst weights and durations to reveal changes in the focus of TKIs research 62 .

4.1Themes involved in TKIs research
We extracted 10,956 unique keywords from the 5,584 available paper titles, and their total frequency was 28,743 (Figure 2).After pre-processing, 6,782 unique words with a total frequency of 26,175 were left.After several rounds of testing and discussion, the threshold value of highfrequency words was taken as 10 in this paper.So after merging the keywords with frequencies lower than 10 into their superordinate words, we finally obtained 296 keywords for the subsequent co-word analysis (Table 1).These 296 keywords whose total frequency accounted for 65.41% (17,120/26,175), can represent the mainstream of TKIs research in the past 5 years and can also reflect a strong concentration trend of TKIs research.

Correlation structure of keywords in TKIs research
Our analysis revealed that the co-word network consisting of the above 296 keywords was exactly the maximal connected subgraph, i.e., none of the high-frequency words in TKIs research is isolated, and all of them have paths associated with others, indicating that the research topics in this field form a whole that is interrelated and interact with each other.

Indicators of the correlation network
The overall indicators of the co-word network are shown in Table 2.The average degree of the network is 49.58, which means that a keyword in TKIs research is directly associated with 49.58 other keywords on average.These 49.58 keywords represent 16.75% of the entire network, which is a relatively small percentage, indicating that the range of inter-topic associations in TKIs research is not extensive.The high degree centralization of the network indicates a strong tendency to be centripetal or concentrated; the high closeness centralization and low betweenness centralization indicates that keywords are most directly related to each other rather than indirectly related; the high clustering coefficient indicates that keywords are likely to cluster into communities with certain words as the core.Collectively, TKIs research has clustered in certain specific sub-directions in recent years, between which the distinction is obvious.However, the density of the current co-word network is not high, i.e., the keywords are not closely related to each other, which indicates that TKIs research is more seriously fragmented and does not form a unified and mature research identity.We further divided the TKIs co-word network into six communities and calculated the module degree 63 to ensure a good division.The network indicators of the keywords reflected their position and role in the TKIs co-word network (Table 3).Non-small Cell Lung Cancer, EGFR, Targeted Therapy, Lung Cancer, EGFR-TKI, Erlotinib, Chemotherapy, Cancer, Sunitinib, and Resistance all appear in the top 10 list of degree centralization and closeness centralization.The research topics associated with these words play an important role and have a strong influence on the whole field, while other words are likely to be clustered into a community with the above words as the core, forming a distinctive research subdirection.In addition to Resistance, the above words also appear in the top 10 list of betweenness centralization, which serves as "bridges" in TKIs research, suggesting that more collaborations or synergies between TKIs research need to pass through these terms.

Analysis of thematic communities
Depending on the association between keywords, TKIs are divided into six clusters or communities with certain important terms at their core.They are C1-Non-small Cell Lung Cancer, C2-Targeted Therapy, C3-Chronic Myeloid Leukemia, C4-HER2, C5-Pharmacokinetics, and C6-ALK (Table 4).These six communities represent the sub-directions of TKIs research in 2016-2020.
They are each closely associated within but loosely associated with each other.
In terms of size, the sub-directions of TKIs research can be divided into 3 echelons (Table 5).The first echelon includes C1-Non-small Cell Lung Cancer, which contains EGFR, EGFR-TKI, EGFR Mutation, Lung Cancer, Erlotinib, etc., and C2-Targeted Therapy, which contains Renal Cell Carcinoma, Sunitinib, Chemotherapy, Cancer, Tumor, etc.These two major sub-directions contain the largest number of keywords and the largest sum of frequencies, which are the main subdirections of TKIs research.The second echelon includes C3-Chronic Myeloid Leukemia, which contains Imatinib, Dasatinib, BCR-ABL, Nilotinib, Gastrointestinal Stromal Tumor, etc., C4-HER2, which contains Apoptosis, Breast Cancer, Lapatinib, Autophagy, Combination Therapy, etc. and C5-Pharmacokinetics, which contains Ibrutinib, Metabolism, Molecular Docking, Plasma, Drug-drug Interaction, etc.These three communities are involved in research topics that are also important for TKIs research.Keywords from these three communities are also important themes in TKIs research.The third echelon only contains C6-ALK, including Crizotinib, Met, Receptor Tyrosine Kinase, ROS1, and ALK-TKI.C6 is still in its infancy, which occupies only a little weight in TKI research.
The varying centrality and density of each community further illustrate that there are sharply differentiated sub-directions in TKIs research (Table 5).For example, C1-Non-small Cell Lung Cancer and C2-Targeted Therapy have the highest centrality, which are the core sub-directions in TKIs research.And C1-Non-small Cell Lung Cancer and C2-Targeted Therapy also have the highest density and are the most developed sub-directions in TKI research.Furthermore, the internal density of each community is higher than the density of the whole TKIs co-word network, which also indicates that each sub-direction is tightly connected internally but loosely connected to each other.The centrality and density of C6-ALK do not have comparative value due to its small size.

Visualization of the correlation network
The diverse correlation structure within and between each sub-direction of the TKIs research is visualized in Figure 4 and Figure 5. Figure 4 shows that there are differences in the influence of varied sub-directions and that the association between these sub-directions is uneven.C1-Non-small Cell Lung Cancer and C2-Targeted Therapy have the strongest associations with the other subdirections, which reflect the strong influence of these two directions in TKIs research.The other four sub-directions are oriented to C1 and C2, or depend on them to varying degrees.The association between C1 and C2 is significantly stronger than that between other sub-directions, so C1 and C2 are the mainstream of current TKIs research.In particular, C1 is the most central and influential subdirection in the whole TKIs research, and the associations between C1 and other directions are generally strong.Isolated C6-ALK is not strongly associated with any other sub-direction except for a closer association with C1.
Figure 5 further shows the correlation structure within each sub-direction, where each term has a different location, function, and role.Each sub-direction has a clear hierarchy, with the most influential terms at the core, which are important research themes, and the more distant from the core of the community, the less important the terms are.For example, EGFR, EGFR-TKI, EGFR Mutation, Lung Cancer, and Erlotinib are important research themes in C1-Non-small Cell Lung Cancer, which are closely related to the other terms or extended to other themes.

The developmentary degree of the sub-directions of TKIs research
Based on Table 5, we draw a strategic diagram (Figure 7) to visualize the developmentary trends of the sub-directions in TKIs research.The C6 community wasn't drawn in the strategy diagram due to its small size and non-comparable network indicators.As shown in Figure 7, we plotted the nodes of different sizes to represent the total frequencies of the varied sub-directions of TKIs research and distributed them in four quadrants according to their density and centrality.C1-Non-small Cell Lung Cancer is in the first quadrant due to its high density and centrality, again indicating that this community is the core direction and most developed in TKIs research.C2-Targeted Therapy is located exactly at the junction of the first and fourth quadrants, which means that C2 is also the core direction of TKIs research, but is in the process of maturing.C3-Chronic Myeloid Leukemia, C4-HER2, and C5-Pharmacokinetics are all in the third quadrant, with relatively low centrality and density, which indicates that they are at the margins of TKIs research and are immature.

Discussion
With the continuous progress of medicine, the targeted drugs TKIs have made great development in basic research and have been widely used for clinical applications.Based on the results of the coword analysis, we have a better understanding of the main research directions of TKIs and accurately assess their maturity, centrality, and interactions.First, in general, TKIs research is unbalanced.Since the first TKIs were introduced, researchers have focused on some hot terms such as NSCLC, EGFR, CML, EGFR-TKI, EGFR Mutation, Erlotinib, Imatinib, Osimertinib, Gefitinib, Targeted Therapy, Renal Cell Carcinoma, Resistance, etc.The above terms can be broadly classified into the following categories: clinical applications (e.g., NSCLC, CML, Renal Cell Carcinoma, Lung Adenocarcinoma), genetic studies (e.g., EGFR, EGFR Mutation, BCR-ABL), typical drugs (e.g., Erlotinib, Imatinib, Gefitinib, Sunitinib, Afatinib, Dasatinib, Osimertinib, Nilotinib), and chemotherapy and drug resistance (e.g., Targeted Therapy, Resistance, Chemotherapy, Drug Resistance).These terms not only reflect the areas of interest of researchers but also indicate the trends of TKIs research.
Based on the visualized co-word network, we found that the research topics tend to be clustered around a few keywords, eventually forming a hierarchical and relatively balanced thematic community.The thematic communities of the TKIs consist of C1-Non-small Cell Lung Cancer, C2-Targeted Therapy, C3-Chronic Myeloid Leukemia, C4-HER2, C5-Pharmacokinetics, and C6-ALK.First, the C1 and C2 communities are the main areas of TKIs research because of their high centrality and frequency.Among them, C1-NSCLC, the largest thematic community, has received a lot of attention from scholars.From gene targets (e.g., KRAS, EGFR, TP53, BIM) to signaling pathways (e.g., IGF-1R, STAT3), from conventional chemotherapy (e.g., Cisplatin, Docetaxel, Pemetrexed) to targeted drugs (e.g., Erlotinib Osimertinib, Gefitinib), scholars have studied NSCLC more and more deeply.C2-Targeted Therapy indicates that TKIs are widely used in targeted therapies, and the application of TKIs has been extended to lung cancer 12,31,32 , breast cancer 64 , renal cancer 65,66 , liver cancer 67 , ovarian cancer 68 , colorectal cancer 69 , leukemia 11,28,29 , thyroid cancer 70,71 , cervical cancer 72 and many other tumors.Chemotherapy regimens containing TKIs were effective in reducing tumor metastasis and recurrence and improving the OS of patients 12,31 .The use of TKIs will be further expanded to more tumor types as more clinical trials are conducted.
Second, C3 and C5 communities have declined during the development of TKIs, both of which are in a marginal position in the strategic diagram.Among C3-Chronic Myeloid Leukemia, Imatinib is the first targeted anti-tumor drug that was first approved by the FDA in 2001 for patients with BCR-ABL positive CML 10,11,28,29,73 .With the emergence of drug resistance, Dasatinib 74 , which targets the SRC, and nilotinib 75 , which targets BCR-ABL, have been applied to resistant patients.Research on CML has been conducted for a long time and this field is now mature, so the application of TKIs has gradually expanded from CML to other diseases, which is leading to the gradual marginalization of the C3 community.C5-Pharmacokinetics is an important interdisciplinary discipline related to TKIs, which is widely involved in the development process of TKIs 76,77 .However, as small molecule drugs, the absorption, transport, distribution, and transformation of most TKIs in vivo have been clearly studied.Meanwhile, several new technologies in molecular biology (e.g., molecular docking 78 and virtual high-throughput screening 79 ) are used more and more, so C5-Pharmacokinetics is gradually fading.
Despite the gradual decline of the C3 and C5 communities, new research areas such as C4-HER and C6-ALK have flourished in recent years.The overexpression, amplification, and mutations of HER2 have been found in a variety of tumors including breast cancer and NSCLC 45 , and targeting HER2 has achieved excellent efficacy in breast cancer 64 .Although several early drugs targeting HER2 had poor efficacy in NSCLC 34,36,45 , the advent of newer generation HER2-targeting drugs such as poziotinib 80 , pyrotinib 81,82 exhibited good anti-tumor effects in clinical trials.Scholars are increasingly interested in targeting HER2 in NSCLC, while research on HER2 for breast cancer is relatively well established.Therefore, C4 may evolve in different directions in the future respectively.The C6-ALK community is small but promising.ALK mutations, especially rearrangements, exhibit strong translational activity in NSCLC 83 .ALK-targeted agents such as alectinib 84 and brigatinib 85 have shown extraordinary efficacy in ALK-positive NSCLC and have become the first-line therapies 86 .Lorlatinib, an ALK inhibitor 87 , appears as a burst word in 2020, and ALK-TKI and ALK Rearrangement have a high weight in Figure 3 from 2019 to 2020, both of which indicate the rapidly rising heat of ALK.In addition, there are still several ALK-targeted drugs in development, so more literature on ALK will be published in the future and the C6 community will grow further.
We found several evolutionary venations, by analyzing the evolution of themes of TKIs research over time.These highly concentrated evolutionary venations indicate scholars' continuous and steady focus on Non-small Cell Lung Cancer, Chronic Myeloid Leukemia, Renal Cell carcinoma, Chronic Lymphocytic Leukemia, etc.These different evolutionary lines show a clear development path: TKIs research is disease-focused and revolved around "gene targets/targeted drugs/resistance mechanisms".For example, in the Chronic Myeloid Leukemia venation, investigators focused on the BRC-ABL in 2016 and focused on dasatinib and nilotinib in 2017-2020, which could both target BCR-ABL and overcome imatinib resistance 74,75 .In the Non-small Cell Lung Cancer venation, investigators focused on EGFR genes in 2016, on EGFR-TKIs represented by osimertinib in 2017-2019, and on resistance mechanisms represented by T790M in 2020.In the Renal Cell carcinoma venation, investigators continued to focus on various TKIs such as sunitinib, sorafenib, regorafenib, and pazopanib, and investigators also paid attention to gene targets such as PDGFR, FGFR, c-Kit, and VEGF, and multidrug resistance.In the Chronic Lymphocytic Leukemia venation, investigators focused on B-cell-derived CLL and lymphoma in 2016, on the aberrant Bruton tyrosine kinase (BTK) from B cells in 2017, and on BTK inhibitors such as ibrutinib that can treat CLL and lymphoma in 2018-2020.Moreover, Lapatinib, a dual EGFR/HER2 TKI 88 , showed good efficacy in breast cancer, which made it an independent evolutionary venation in 2017 with continues heat to now.
In addition to the main few evolutionary venations, we identified some isolated themes that depict the current state of TKIs research.Crizotinib gained attention as an ALK inhibitor in 2016, but Crizotinib's declined rapidly due to its poor efficacy and the emergence of second-generation ALK inhibitors, making Crizotinib & ALK an isolated topic.Circulating Tumor DNA (ctDNA) is important for the efficacy assessment and prognosis analysis of tumors.And it's the future trend in TKIs research to use NGS or liquid biopsy technology to precisely analyze ctDNA in cell-free DNA.The PI3K/AKT pathway plays an important role in cell growth, proliferation, migration, and angiogenesis, which can be activated by RTKs.So as one of the mechanisms of TKIs, the apoptosis & PI3K/AKT venation was noticed in 2017-2018.Cabozantinib is a multi-target TKIs that can target 9 genes (e.g., AXL, Met) 89 , and Cabozantinib & AXL appeared as a separate topic in 2020.Furthermore, the burst words pyrotinib and anlotinib, which accounted for a relatively large weight in 2020, are also multi-target inhibitors 81,82,90 .This suggests that multi-target drugs may become an important direction for the development of TKIs and will likely receive more attention in the future.Immune checkpoint inhibitors and TKIs are important drugs for tumors.It's discovered that the combination of PD-L1 inhibitors (e.g., Pembrolizumab) and TKIs (e.g., Lenvatinib) in patients with malignant tumors was more effective than single drug 91 , suggesting that the combination therapy was an important development direction for future tumor therapy.For well-known reasons, many patients being treated with TKIs were co-infected with SARS-COV-2 in 2020 92 , and it was also suggested that some TKIs such as BTKs may have therapeutic effects on COVID-19 93 , which made COVID-19 a burst word in TKIs research.As the global epidemic of COVID-19 in 2021, investigators' interest in TKIs for patients infected with SARS-COV-2 would further increase.

Conclusions
In conclusion, we revealed the hotspots, knowledge structure, and trends of evolution of TKIs research during 2016-2020 utilizing co-word analysis and literature visualization.Based on the above results, we speculate that the general status of TKIs research is as follows: (1) NSCLC and CML are the most important clinical application areas for TKIs; (2) EGFR is the most common target gene for TKIs, and EGFR-TKIs are the most commonly used molecular-targeted TKIs, among which Erlotinib, Osimertinib, and Gefitinib have gradually matured; (3) TKIs have become a mature field for targeted therapeutic applications, and drugs targeting HER2 and ALK has further expanded the application of TKIs; (4) drug resistance remains a major challenge for TKIs.Of course, our study also has some limitations: first, our search included only English literature from 2016 to 2020, while non-English literature was excluded; second, the co-word analysis didn't take the quality, influence, and rigor of the literature into account, which was a common shortcoming of such papers [94][95][96] .In a nutshell, our work remains valuable in revealing the knowledge structure and evolutionary trends of TKIs research.

Figure 1 .
Figure 1.Search procedure for documents in TKIs research.DE: descriptor; TI: title; WOSCC: Web of Science Core Collection.

Figure 2 .
Figure 2. Yearly number of papers and words related to TKIs research (2016-2020)

Figure 3 .
Figure 3. Burst disciplines of TKIs research from 2016 to 2021 Burst keywords can represent important changes in TKIs research.Figure 3 showed a varying number of burst words in TKIs research each year, whose duration expressed in terms of the length of the horizontal bar and weight in terms of the area.As can be seen from figure 3, a variable number of emergent terms have appeared in TKIs research every year since 2016, especially in 2019 and 2020, indicating the emergence of new research themes in this field every year.And the greater weight of burst words in 2020-2021 (e.g., HER2, Pyrotinib, Next-generation Sequencing, COVID-19, Immunotherapy, ALK-TKI, ALK Rearrangement, Cell-free DNA ， Liquid Biopsy，Personalized Medicine), suggests that these words in TKIs research were extensively explored by researchers in 2020.

Figure 4 .
Figure 4. Correlation structure of sub-directions in TKIs research

Figure 5 .
Figure 5.The internal correlation network structure of each sub-directions in TKIs research 4.3 Evolution patterns of and trends in TKIs research

Figure 7 .
Figure 7.The relative development status and trends of 5 sub-directions in the strategic diagram.

Table 1 .
Top 100 keywords in papers related to TKIs research

Table 2 .
The whole network indicators

Table 3 .
Top-10 keywords in terms of degree, betweenness, and closeness centrality

Table 4 .
Topic communities related to TKIs research

Table 5 .
Indicators of 6 theme communities in TKIs research.