MeSH terms attached to articles published between 1996 and 2020 were collected through PubMed (https://www.ncbi.nlm.nih.gov/pubmed/) on August 17th, 2020. A total of 21,706,508 articles were included in the analysis.

MeSH terms attached on each article were identified from the XML data, and any overlaps in terms for each article were eliminated by our original Perl scripts as described in our previous articles (Ohniwa et al., 2010;Ohniwa and Hibino, 2019). Here, to identify the set of MeSH terms attached to each article, terms tagged as <DescriptorNameUI = @> and <NameOfSubstance UI = @> from the XML data (@ represents each UI) were extracted, and then overlaps in the terms for each article were eliminated by Perl with our original scripts. Then, to eliminate the terms not concerned with research topics, all the MeSH terms under the following hierarchies were excluded: “Geographicals [Z],”“Publication Characteristics [V],”“Named Groups [M],”“Health Care [N],”“Information Services [L01.453],”“Communications Media [L01.178],”“Communication [L01.143],”“Information Centers [L01.346],” and “Publishing [L01.737]” according to the 2018 MeSH Tree Hierarchy (https://meshb-prev.nlm.nih.gov/search). These categories involve MeSH terms which represent the style or the background of articles rather than its research contents. A total of 1,776,759 kinds of terms with a total of 957,790,657 occurrences were obtained between 1996 and 2020 by this operation.

To identify the sets of authors and affiliations attached to each article, terms tagged as <AffiliationInfo>, <LastName>, and <Initials> within <Author ValidYN = @> (@ represents each “Y or N”) were extracted by our original Perl script. A total of 42,923,027 kinds of authors (with affiliation) with a total of 56,467,393 occurrences were obtained by this operation (in the case of only author name, a total of 7,395,577 kinds of authors with a total of 108,977,113 occurrences).

Coronavirus-related articles published between 1996 and 2020 were separately obtained by searching PubMed using the query terms of “coronavirus OR COVID-19 OR SARS-CoV OR MERS-CoV.” A total of 56,077 articles and 66,850 kinds of MeSH terms with a total of 1,739,841 occurrences were obtained between 1996 and 2020 by this operation. From these articles, a total of 252,292 kinds of authors (with affiliations) with a total of 273,759 occurrences were obtained (in the case of using only authors’ names, a total of 157,593 kinds of authors with a total of 337,946 occurrences).

The number of articles having the following words in the affiliation was counted: “Argentina”, “Australia”, “Austria”, “Belgium”, “Brazil”, “Bulgaria”, “Canada”, “Chile”, “China (‘China’ or ‘People’s Republic of China’)”, “Croatia”, “Czech Republic”, “Denmark”, “Egypt”, “Finland”, “France”, “Germany”, “Greece”, “HongKong”, “Hungary”, “India”, “Iran”, “Ireland”, “Israel”, “Italy”, “Japan”, “Korea”, “Malaysia”, “Mexico”, “Netherlands”, “New Zealand”, “Norway”, “Pakistan”, “Poland”, “Portugal”, “Romania”, “Russia”, “Saudi Arabia”, “Singapore”, “Slovakia”, “Slovenia”, “South Africa”, “Spain”, “Sweden”, “Switzerland”, “Taiwan (‘Taiwan’ or ‘Republic of China’)”, “Thailand”, “Turkey”, “Ukraine”, “United Arab Emirates (‘UAE’ or ‘United Arab Emirates’)”, “United Kingdom (‘England’, ‘U.K’, ‘UK’, ‘United Kingdom’, ‘Scotland’)”, and “United States (‘USA’ or ‘United States’)”.

Among MeSH terms, we arbitrary defined unique keywords to coronavirus research as follows: ( A α   in   β C / B β ) / ( C α   in   β / D β ) ≥ 2 , where A_{α in β} is the number of appearances of the MeSH term α in years β found in coronavirus-related articles, B_β is the total number of the coronavirus-related articles in years β, C_{α in β} is the number of appearances of the MeSH term α in years β in PubMed, and D_β is the total number of articles counted in years β in PubMed. The terms whose rates were more than 2 were defined as unique keywords. A total of 13,125 kinds of MeSH terms as unique keywords with a total of 1,739,841 occurrences were collected between 1996 and 2020.

Top 50 most frequently occurring unique keywords were collected for each period, and they were examined whether they coappeared in the same article. The coappearance was examined by using Perl with our original scripts. The coappearance of the keywords was visualized using Pajek software (Batagelj and Mrvar, 2002). To eliminate any weak relation among keywords, the threshold for making edges was set at 10% of the number of keywords (selecting smaller sized nodes) linked by the edges, according to the clusters appeared in the networks visualized by Pajek.

Regarding their impact on research contents, this study identified unique keywords which represented the characteristics of coronavirus research before and after the outbreaks of SARS, MERS, and COVID-19. Since the collection of frequently appeared MeSH terms by itself did not reveal the unique characteristics of the research contents owing to their generality of use in numerous articles (Ohniwa et al., 2010), unique keywords from coronavirus research in a particular year were selected. These unique keywords defined as MeSH terms in coronavirus-related articles had the appearance rate that was at least twice as high as their appearance rate in all articles of the year (see Materials and Methods). This operation helped identify representative terms such as “SARS Virus” between 2003 and 2006 and “Middle East Respiratory Syndrome Coronavirus” between 2013 and 2016 as top 25 frequently appeared unique keywords (Table 1). Keywords related to “biological classification of coronaviruses,”“infection matters of coronaviruses,” and “respiratory tract issues [respiratory tract is the target infection site for coronaviruses as well as the site where its major symptom appears (Channappanavar and Perlman, 2017;Singhal, 2020)]” were commonly used between 1996 and 2020—consistent with the collection of unique keywords from coronavirus-related articles in this study. Keywords related to “components of the virus” and “biological aspect” were found in all of the years except for 2020, while others such as “public health” and “human” began to appear in 2020. When the list was expanded to top 50 frequently appeared unique keywords, this tendency was even strengthened (Supplementary Table S1). In the meantime, conventional keywords such as “immunology” were still found in all of the years. In addition to the fact that these top 50 keywords frequently coappeared in the coronavirus-related articles (Supplementary Figures S1–S6), the results of this investigation suggest the following: 1) the regional outbreaks of SARS and MERS did not change the cross-disciplinary research trends that had existed before the outbreaks as keywords fell well within the scope of existing well-connected networks for coronavirus research focusing on the topics which had been constantly studied regardless of the outbreaks; 2) the global outbreak of COVD-19, on the other hand, initially had a diversifying impact on the existing research trend as newly emerged keywords formed unconnected research networks across different disciplines, including areas of research such as jurisprudence and public policy.

The increase in the number of coronavirus-related articles after the outbreaks of SARS, MERS, and COVID-19 was largely due to the entry of new researchers (Supplementary Table S2). For example, between 2015 and 2019, a total of 24,745 authors with specific affiliations were identified in coronavirus-related articles. Their names with the same affiliations were found in only 369 out of 39,804 (0.9%) coronavirus-related articles in 2020. In the case of SARS, the authors of publications between 1998 and 2002 accounted for 23 out of 2,805 (0.8%) published articles on coronavirus between 2003 and 2006. In the case of MERS, the authors from the years between 2008 and 2012 occupied 123 out of 3,052 (4.0%) coronavirus-related articles published between 2013 and 2016. As a reference, we also analyzed the case of “Influenza,” resulting in a higher rate of occupation [the authors from the years between 2014 and 2018 occupied 1,364 out of 5,542 influenza-related articles published in 2019 (24.6%)]. Here, because not all affiliations were attached to the authors in PubMed before 2014 (Supplementary Table S3), the authors regardless of their affiliations were counted in (Supplementary Table S4). The risk of counting different authors as the same authors had to be taken. In any case, the results of this investigation show that most of these converted experts were conducting their coronavirus research on a temporary basis as they did not continue publishing on coronaviruses after new outbreaks.

This study then examined unique keywords used by the newcomers to coronavirus research from the past 5 years (searched by their names with affiliations). Those listed as top 25 frequently appeared unique keywords used by the newly entered researchers (Table 2) did not largely overlap with those of coronavirus research between 1996 and 2020 (Table 1) (11 out of 72 keywords). This tendency did not change when we compared them with top 50 frequently appeared unique keywords (Supplementary TablesS1, S5). Thus, it is likely that such new researchers came from different disciplines.

After the outbreak of SARS, experts on “infection of RNA viruses and bacteria especially in the fields of immunology, molecular biology, bioinformatics, and/or sociology” began to take part in coronavirus studies (Figure 2A). In the case of MERS, many RNA virus researchers handling “influenza virus infection” and “HIV infection” as well as experts on “health-care issues” joined in (Figure 2B). In the case of COVID-19, it attracted experts on the “hepatitis virus” and “mycobacterium” (Figure 2C).

Furthermore, for COVID-19, many converted experts had no background in the research on “RNA virus and infection.” They consequently formed a separate scientific network apart from those with their background in “RNA viruses and infections” in the beginning. They started from forming networks with those with their background in “urology and prostate,”“diagnosis of digestive system,”“clinical trial,” or “health care and guide line planning,”“jurisprudence,”“public policy,” and others. Such outside-in networking, due to the wide disciplinary base of the network, was one of the characteristics of COVID-19 research dynamics in the beginning.

Once a novel infectious disease emerges in a certain country, scientific publications on the disease increase in the corresponding country (Table 3 and Supplementary Table S6). The United States and Germany have always been ranked among top 10 countries frequently publishing coronavirus articles, regardless of contingencies involving novel coronaviruses—suggesting that these countries have been the leaders of coronavirus research over the past 25 years. China became the No. 2 country after the SARS outbreak in November 2002 in China. Taiwan, Hong Kong, Singapore, and Canada—the countries and regions that prevailed after being hit hard by SARS in 2003 (Chan-Yeung and Xu, 2003; Wallinga and Teunis, 2004)—were also in top 10 between 2003 and 2006. In the case of the MERS outbreak in November 2012, Saudi Arabia was ranked among top 10 as one of the major countries that overcame the epidemic. Korea was also ranked among top 10 countries between 2013 and 2016, quite possibly due to the outbreak of the MERS epidemic in Korea in 2015 (Chen et al., 2017). With regard to the case of COVID-19, the exponential increase in the number of scientific publications originating from all the countries examined in this article suggests its impact on a global scale. It was also found that many researchers converted to the field of coronavirus studies after the outbreaks of infectious diseases (Supplementary Table S7). Accordingly, the outbreaks of novel coronavirus diseases accelerate coronavirus research activities in the affected countries by attracting new researchers.

The proportion of internationally co-authored articles on coronaviruses to all the coronavirus-related articles among the countries listed in Supplementary Table S6 was 0.22–0.26 between 2016 and 2019 (Supplementary Table S8). This rate was higher than that of all the internationally co-authored articles during the same period (0.16–0.18). Thus, compared with the average collaboration ratio of articles, a higher rate of transnational collaboration was present for coronavirus research. In contrast, in 2020 after the outbreak of COVID-19, the rate of internationally co-authored coronavirus articles was 0.16. This was lower than the rate of all the international research articles published in the same year (0.19). In this way, transnational collaboration on coronavirus research as indicated in the rate of international co-authorships has decreased since the outbreak of COVID-19.