Mapping the Emergence of Synthetic Biology

Abstract

In this paper, we apply an original scientometric analyses to a corpus comprising synthetic biology (SynBio) publications in Thomson Reuters Web of Science to characterize the emergence of this new scientific field. Three results were drawn from this empirical investigation. First, despite the exponential growth of publications, the study of population level statistics (newcomers proportion, collaboration network structure) shows that SynBio has entered a stabilization process since 2010. Second, the mapping of textual and citational networks shows that SynBio is characterized by high heterogeneity and four different approaches: the central approach, where biobrick engineering is the most widespread; genome engineering; protocell creation; and metabolic engineering. We suggest that synthetic biology acts as an umbrella term allowing for the mobilization of resources, and also serves to relate scientific content and promises of applications. Third, we observed a strong intertwinement between epistemic and socio-economic dynamics. Measuring scientific production and impact and using structural analysis data, we identified a core set of mostly American scientists. Biographical analysis shows that these central and influential scientists act as "boundary spanners," meaning that their importance to the field lies not only in their academic contributions, but also in their capacity to interact with other social spaces that are outside the academic sphere.

Fig 1. Global population statistics of the SynBio community over time.

Cumulated number of publications (bar chart), newcomers’ ratio and largest component relative size. While the number of publication follows a typical exponential growth, the newcomers ratio, while very high, is decreasing with time, and the largest component relative size has been significantly growing since 2010 indicating a progressive structuration of the SynBio community.

Fig 2. Phase diagram plotting the connected component relative size according to the cumulated number of authors at different years.

Zebrafish and BRCA scientific communities were also plotted for the sake of comparison. SynBio (SB blue crosses) exhibits less “structuration” than those other fields for the same population size, but the gap is rapidly decreasing.

Fig 3. References co-citation map.

The 100 most cited references are considered to build a co-citation network. Node sizes scale with the number of citations received by references. Node color depends on their cluster assignment. A colored circle is also drawn at the barycenter of each cluster which size is proportional to the number of related papers (that is the number of papers citing a significant number of references from this cluster). Edge widths are proportional to the strength of the link between references. Cluster labels (bold and upper-case) were manually added. During this labelling process, we gathered 5 very tightly connected clusters under the same family.

Fig 4. Semantic map showing the structure of 200 most pertinent terms extracted from titles, abstracts and keywords in the corpus.

While some semantic clusters seem to be clearly resonant with co-citation clusters (Artificial cells/ (3) protocol creation), others are more instrumental or seem more conceptual (e.g. gene expression optimization, system biology).

Fig 5. Contingency matrix between co-citation and semantic clusters.

Each cell in the matrix shows how correlated/uncorrelated/anti-correlated a co-citation and semantic cluster are, meaning that the number of papers assigned to both clusters is higher/lower than the expected value we could expect from a random model. Red cells code for a positive deviation, while blue cells indicate a negative deviation. The intensity of colors shows the intensity of the deviation. For instance the deepest red cell shows that there are more than five times more papers assigned to both “(3) Protocell creation” and “Artificial Cells” than expected if co-citation and semantic cluster distributions were independent, conversely “Protocell creation” is anti-correlated with “Mammalian cells” (Five times fewer papers than expected. Crosses indicate cells where the deviation from the null model is not statistically significant).

Fig 6. Position analysis of the main actors of SynBio community according to their centrality and impact.

The joint distribution of nodes betwenness centrality and cumulated Impact is represented through the hexagonal heatmap showing how concentrated the population is: more than 10000 scientists (over 11321) actually have both a very low impact and centrality. Conversely, the 30 most important scientist (according to the product of impact and centrality) are plotted with a pie diagram showing their paper distribution among the 4 different approaches (green for Biobricks engineering, yellow for genome engineering, purple for metabolic engineering and pink for protocell creation). The epistemic landscape is strongly defined by members of this core-set: they have indeed co-authored 77 over the 100 most cited articles by the corpus.