Progress and Knowledge Transfer from Science to Technology in the Research Frontier of CRISPR Based on the LDA Model

The prokaryote-derived CRISPR (clustered regularly interspaced short palindromic repeats) is a genetic engineering technique in molecular biology by which the genomes of living organisms, including humans, may be modified (Ledford, 2015). It enables targeted genetic modifications in cultured cells, as well as whole animals and plants with extremely high precision, cheaply, and with ease (Doudna, 2020; Kim & Kim, 2014; Zhu, Li, & Gao, 2020). Especially, the CRISPR/Cas9 genetic scissors discovered in 2012 by Emmanuelle Charpentier and Jennifer A. Doudna who won the 2020 Nobel Prize in Chemistry have not only revolutionized basic science but also resulted in innovative crops and will lead to ground-breaking new medical treatments (http://www.nobelprize.org/). The diversity, modularity, and efficacy of CRISPR-Cas systems are driving a biotechnological revolution and the field of CRISPR-based biotechnology is developing at a rapid pace (Doudna, 2020; Doudna & Charpentier, 2014; Pickar-Oliver & Gersbach, 2019).

The impact of recent development of CRISPR on science and biotechnology is immense (Doudna & Gersbach, 2015). Academic research in this area has made great progress especially since its breakthroughs in 2012 (Knott & Doudna, 2018). A few investigations aimed at assessing the development and status of this field have been conducted, for example, Qin, Wang, and Ye (2019) conducted a research to reveal the main research hospots of CRISPR/Cas9 by constructing and analyzing core keyword co-occurrence networks. Zhou et al. (2021) studied the evolution of academic research hotspots in CRISPR by analysing the usage of key phrases. However, most of them are limited to general bibliometric analysis based on a small amount of selected data. Today, an increased interest has been shown in using topic modeling methods to explore the structure of a certain domain (Figuerola, Marco, & Pinto, 2017; Han, 2020; Lamba & Madhusudhan, 2019; Sugimoto et al., 2011). The method presents a broader and more comprehensive perspective based on a large number of publications.

Science research is also regarded as a driving force for technology innovation and economic growth (Gittelman & Kogut, 2003; Li, Azoulay, & Sampat, 2017; Lo, 2010). New basic knowledge may lead to a knowledge transfer to technological fields resulting from interactions of citations, and, in this way, promotes innovations in industry and progress in society (Hu & Rousseau, 2018). Besides, this incredible technique over our heritable information, CRISPR, also brings unknown risks to human beings. Thus, it becomes very important to track and monitor the technological applications of CRISPR, which can also help to ensure responsible innovations by scientists and use it in an ethical and safe way while driving the technology forward (Baltimore et al., 2015; Gurwitz, 2014).

In this article, we present a new perspective on observing and understanding the research topics and their development in CRISPR based on the LDA model and try to illustrate the knowledge transfer trends from research topics to technological fields based on paper-patent citations.

Publications covering the genome editing toolbox based on CRISPR systems such as Cas9, Cas12, Cas13, base editors (BEs), and prime editors (PEs) were drawn from Clarivate Analytics’ Web of Science (WoS) Core Collection in August 2021. Specifically, we used the following search string.

TS= (“CR ISPR*” OR “clustered regularly interspaced short palindromic repeats*” OR “base editor*” OR “base editing” OR “prime editing” OR ((Cas9 OR Cas10 OR Cas11 OR Cas12 OR Cpf1 OR Cas13 OR Cas14 OR CasX OR gRNA OR sgRNA) AND (“gene editing” OR “genome engineering” OR “genome editing” OR “genome editor” or “genome binding”)))

First, we performed data cleanup and removed irrelevant data. Final data were limited to the article type, and included publications with a title, an abstract, keywords and the publication year. This totaled 15,904 records between 2011 and 2020. The title, keywords and abstract text were saved in a Microsoft Excel file for the LDA model, and the publication year was used to incorporate the LDA results for topic changes analysis.

Next, we searched by article title on lens.org to trace their patent citations one by one. For articles that have been cited by patents, we downloaded their citing patents and assigned them a unique identifier to link them to their citing patents. In this way we obtained for each article the number of patent citations, the application year, and the corresponding International Patent Classification codes (IPC codes) for each patent. Note that, we only counted the number of different families to avoid double counting. Totally, we obtained 18,985 citing patent families of all articles. Then we developed a Python script to get the first four digits of the IPC code, which is the IPC-4 code dividing the patent into a subclass level (WIPO, 2019). After that, we constructed a matrix of relationships between scientific topics and IPC-4 codes (based on their citation links). Here, only different IPC-4 codes were counted. Moreover, Python, Microsoft Excel, Gephi, and Tableau were applied to data visualization.

In this contribution, we applied the LDA model for topic detection in the transformative area, CRISPR, and demonstrated the development of topics over time. Results show that the dominant topics have gradually evolved from “Mechanism of CRISPR/Cas” to “Gene mutation based on CRISPR/Cas9” and “Genome editing based on CRISPR system”. Moreover, by tracking all patent citations of articles, we discovered that the publications in this area are highly technology relevant and affect the technological world at an extremely rapid pace. Technological influence varies with research topics, among which the topics “Mechanism of CRISPR/Cas”, “Biotechnology with CRISPR/Cas9” and “Genome editing based on CRISPR system” form top 3. Finally, as shown in Fig. 5, we figured out a big picture of knowledge transfer between S&T on CRISPR, where the C12N and A61K domains are the two most important inflows from each topic.

As a transformative technique, CRISPR attracted considerable attention from worldwide scientists and international organizations. Although many experts have described the development and applications of CRISPR from the perspective of qualitative analysis, such as recapitulating genetic mutations in animals or cellular models (Hsu, Lander, & Zhang, 2014), programmable RNA targeting and viral gene disruption (Doudna & Charpentier, 2014), DNA changes in stem cells and treating human diseases (Baltimore et al., 2015), yet they just focused on limited aspects of CRISPR in a traditional mode of review (Hsu, Lander, & Zhang, 2014; Doudna & Charpentier; 2014; Baltimore et al., 2015). In contrast, our contribution provided more detailed and objective results including the research topics in CRISPR and their development based on a large number of publications in the past decade. Hence, a full picture of the new technology was taken up, which also helps scholars grasp the current research trends and seize the opportunity of scientific research on CRISPR.

The pathway of knowledge transfer is a main bridge for understanding the interaction between science and technology. It is seen as an essential source of innovation and a mechanism for the dissemination of research results (Campbell et al., 2020; Wang & Ye, 2021). By measuring knowledge transfer through direct paper-patent citations in the frontier of CRISPR, we found that scientific publications and their impact on the technological field vary by research topic. The results not only provide meaningful clues for knowledge management in this domain, but also benefit policy-makers in formulating scientific and technological policies, as well as allocating scientific resources. For example, the SKT value of “Biotechnology with CRISPR/Cas9” is 0.31, while SKT values of “CRISPR in agriculture”, “Targeted therapies” and “Genome analysis on strains” are only 0.01. The largest knowledge flow (SKT value of 0.1493) is from the topic “CRISPR/Cas9 biotechnology” to technological field “C12N”. The findings above are valuable for the decision-making of research foundations, e.g. funders can decide whether to support the technology-relevant topics or the basic-oriented topics according to the scientific or economic development goals.

The rapid development and broad application prospects of CRISPR also come with a great responsibility to use it ethically and safely (Doudna & Gersbach, 2015). This simple and widely available technology can now be used to perform genome modification in eggs, sperm or embryos, thereby changing the genetic makeup of human germline, human beings are facing unknown risks in science, medicine, and ethics (Doudna, 2020; Baltimore et al., 2015). Thus, the application of this technology, especially topics on “Human therapeutic” and “Genetic engineering human cells” in the above results, must be rigorously regulated. Our research provides a good reference for regulators in tracking the technological application of CRISPR to drive the technology forward while ensuring responsible use.

Many studies have confirmed that LDA can effectively cluster meaningful and interpretable topics from a large number of documents (Blei & Lafferty, 2007; Suominen & Toivanen, 2016; Yau et al., 2014). In today's world, various extensions of LDA have been proposed to further detect topic changes over time, such as the Topic Over Time model (TOT) and the Dynamic Topic Model (DTM) (Blei & Lafferty, 2006; Shan & Li, 2010; Wang & McCallum, 2006). As a transformative tachnology, the publications on CRISPR has been explosively growing in the latest decade (as shown in Fig. 1). Hence in this study, we applied the time post-discretized analysis to detect the topic changes, on the basis of the number of articles produced each year (Griffiths & Steyvers, 2004; Shan & Li, 2010), that is, running the LDA model based on the entire text set and then incorporating the publication year of texts into the LDA results. This method has been serving as a reliable means to gain insight into the dynamics of science (Figuerola, Marco, & Pinto, 2017; Griffiths & Steyvers, 2004; Shan & Li, 2010; Jiang et al., 2021). The result of this work may be beneficial for scientists, intelligence analysts and policy-makers to meet the challenges related to this disruptive technology.

Our study has limitations. We only focused on publications related to CRISPR retrieved from the Web of Science and their citing patents indexed in lens.org. Besides, a limitation inherent with LDA analysis is in the manual interpretation and labeling of “topics”. Some topics are fairly straightforward to label by interpreting the word distribution and examining the most representative articles of them in detail, while others are proved more difficult to ascertain the content or relationship that connected the words and articles. We hope that further research will provide additional insights built on our work.