Science and Technology Relatedness: The Case of DNA Nanoscience and DNA Nanotechnology

See the discussion on neo-classical and evolutionary theories in Nelson and Winter (1974) and concerns raised by Dosi (1982), Suenaga (2015), and others about uncertainties related to S&T that may cause new technological paradigms.

We later refer to these as PPPs.

We later refer to it as NPL.

The main path approach is a network analysis tool introduced in the late 1980s to investigate networks of scientific publications, and later to study patent networks (see Verspagen, 2007; Bekkers & Martinelli, 2012). The top main path is considered as representing the most important developments in citation networks.

In a similar vein, Heinisch et al. (2016) used co-location as a proxy for direct knowledge interaction.

Both directly and indirectly.

Note that while we use the term “academic publications,” such publications can also be authored by people working for firms. Likewise, university staff can also apply for patents.

They are “term groups,” which consist of synonyms, abbreviations…which have the same meaning.

We used two levels of analysis: domain level, and term level.

This may due to the fact there is no fixed perfect definition for a new field.

Precision can be estimated by taking a random sample of the set, and manually investigating whether all the records indeed belong to the sought field. Recall can be estimated by independently creating a set of records that are known to belong to the sought set (e.g., by asking an independent expert in the field, or selecting the relevant patents or publications of key contributors) and then testing whether these records are present in the set.

We found that, in our context, four was the number of concepts allowing us to reach the best balance between recall and precision. With three concepts, the level of precision reduced significantly. With five concepts, the concepts started to lose their initial independence, and the level of recall dropped.

Information sources include materials and notes taken at technical conferences on DNA-Nano, communication with experts by email and Skype, and publications and news items in the field of DNA-Nano.

Sungi Kim, PhD candidate at Seoul National University, validated the queries for collecting publications. Jürgen Schmied, CEO of Gattaquant, a company working in the field of DNA Nanotechnology, validated the queries for collecting patents.

We improved recall by checking whether the authors and inventors whom we know are present in our search results. If not, we included more keywords from their publications/patents. We improved precision by sampling 20 records each time and checking if any record is irrelevant. Then we identified the keywords that distinguish DNA-Nano from other fields in that record, and put them in the exclusion terms.

And even those with high term frequency-inverse document frequency (tf*idf).

For instance, “this study,” “this invention.”

We ended up with 109 cross-domain term groups, which have been harmonized from 400 technical terms extracted with highest scores by the automatic Term Recognition algorithm.

We used Higuchi Koichi’s KH coder text-mining software (Version 3a12d).

A sub-domain unit of analysis.

We used Stephan Evert’s R package “corpora” for this specific Chi-square test.

We explained how we selected 109 term groups in Sect. 3.2. For the actual analysis of S&T relatedness, we called them “terms” for convenience.

This first step resulted in 538 pairs in Science and 391 pairs in Technology.

This second step resulted in 133 pairs of direct complementarity and 10,525 pairs of indirect complementarity.

In earlier ages, however, the temporal relatedness between S&T could happen in 2000 years (Johns, 2020).

These robustness checks are available upon demand from the authors.

This does not happen with knowledge areas that are neither similar nor complementary.

We did so at the third workshop on Functional DNA Nanotechnology (6–8 June 2018, Rome, Italy).

We explained the concepts of direct and indirect complementarity, and gave them the list of 133 pairs of terms. Some experts reacted right away, others responded later by email.

For reasons indicated in Sect. 4.2, we did not measure indirect complementarity over the different subperiods.

It is worthwhile noting that some records where the concept “nanotechnology” is implicit, should be included in our datasets. Certain inventors choose not to mention nano-related terms explicitly or discuss only DNA or oligonucleotides. That might be the reason why a considerable number of patents belonging to DNA Nanotechnology is not classified under IPC-code B82 (Nanotechnology). From a conceptual point of view, DNA and nano are quite different. However, from a practical point of view, when discussing DNA or nucleotides, we should imply that the research is conducted at the nanoscale, since the dimension of a DNA strand is approximately 2.5 nm. Therefore, “DNA” and “*nucleotid*” are included in two concept areas (Nanotechnology and DNA) to avoid missing certain records that do not mention nano-related terms. Although DNA and Nanotechnology are closely related concepts, we have not grouped them because this leads to considerably more noise in the datasets selected. Thus, DNA-related terms must appear in the set under any conditions, while the presence of nano-related terms remains an option.

It is harder practically to achieve precision in retrieving patents rather than retrieving publications. Therefore, we decided to adjust terminologies in this DNA concept group into more specific terms, which include DNA.