Search

Fundamental scientific advances can take decades to translate into improvements in human health. Shortening this interval would increase the rate at which scientific discoveries lead to successful treatment of human disease. One way to accomplish this would be to identify which advances in knowledge are most likely to translate into clinical research. Toward that end, we built a machine learning system that detects whether a paper is likely to be cited by a future clinical trial or guideline. Despite the noisiness of citation dynamics, as little as 2 years of postpublication data yield accurate predictions about a paper’s eventual citation by a clinical article (accuracy = 84%, F1 score = 0.56; compared to 19% accuracy by chance). We found that distinct knowledge flow trajectories are linked to papers that either succeed or fail to influence clinical research. Translational progress in biomedicine can therefore be assessed and predicted in real time based on information conveyed by the scientific community’s early reaction to a paper., Fundamental scientific advances can take decades to translate into improvements in human health. This study shows that a machine learning model can accurately predict whether an article is likely to be cited by a future clinical trial or guideline, using as little as two years of post-publication citation data.

Scientific research is often thought of as being conducted by individuals and small teams striving for disciplinary advances. Yet as a whole, this endeavor more closely resembles a complex and integrated system of people, papers, and ideas. Studies of co-authorship and citation networks have revealed important structural properties of researchers and articles, but currently the structure of scientific ideas themselves is not well understood. In this study, we posit that topic networks may be a useful framework for revealing the nature of conceptual relationships. Using this framework, we map the landscape of interconnected research topics covered in the multidisciplinary journal PNAS since 2000, constructing networks in which nodes represent topics of study and edges give the extent to which topics occur in the same papers. The network displays small-world architecture, characterized by regions of dense local connectivity with sparse connectivity between them. In this network, dense local connectivity additionally gives rise to distinct clusters of related topics. Yet notably, these clusters tend not to align with assigned article classifications, and instead contain topics from various disciplines. Using a temporal graph, we find that small-worldness has increased over time, suggesting growing efficiency and integration of ideas. Finally, we define two measures of interdisciplinarity, one of which is found to be positively associated with PNAS's impact factor. Broadly, this work suggests that complex and dynamic patterns of knowledge emerge from scientific research, and that structures reflecting intellectual integration may be beneficial for obtaining scientific insight.

The ever-increasing quantity and complexity of scientific production have made it difficult for researchers to keep track of advances in their own fields. This, together with growing popularity of online scientific communities, calls for the development of effective information filtering tools. We propose here an algorithm which simultaneously computes reputation of users and fitness of papers in a bipartite network representing an online scientific community. Evaluation on artificially-generated data and real data from the Econophysics Forum is used to determine the method's best-performing variants. We show that when the input data is extended to a multilayer network including users, papers and authors and the algorithm is correspondingly modified, the resulting performance improves on multiple levels. In particular, top papers have higher citation count and top authors have higher h-index than top papers and top authors chosen by other algorithms. We finally show that our algorithm is robust against persistent authors (spammers) which makes the method readily applicable to the existing online scientific communities.

Background As statisticians develop new methodological approaches, there are many factors that influence whether others will utilize their work. This paper is a bibliometric study that identifies and quantifies associations between characteristics of new biostatistics methods and their citation counts. Of primary interest was the association between numbers of citations and whether software code was available to the reader. Methods Statistics journal articles published in 2010 from 35 statistical journals were reviewed by two biostatisticians. Generalized linear mixed models were used to determine which characteristics (author, article, and journal) were independently associated with citation counts (as of April 1, 2017) in other peer-reviewed articles. Results Of 722 articles reviewed, 428 were classified as new biostatistics methods. In a multivariable model, for articles that were not freely accessible on the journal’s website, having code available appeared to offer no boost to the number of citations (adjusted rate ratio = 0.96, 95% CI = 0.74 to 1.24, p = 0.74); however, for articles that were freely accessible on the journal’s website, having code available was associated with a 2-fold increase in the number of citations (adjusted rate ratio = 2.01, 95% CI = 1.30 to 3.10, p = 0.002). Higher citation rates were also associated with higher numbers of references, longer articles, SCImago Journal Rank indicator (SJR), and total numbers of publications among authors, with the strongest impact on citation rates coming from SJR (rate ratio = 1.21 for a 1-unit increase in SJR; 95% CI = 1.11 to 1.32). Conclusion These analyses shed new insight into factors associated with citation rates of articles on new biostatistical methods. Making computer code available to readers is a goal worth striving for that may enhance biostatistics knowledge translation.