Career destiny. A statistical model based on the publication record of more than 25,000 researchers finds that science really is "publish or perish."

John Bohannon

Career destiny. A statistical model based on the publication record of more than 25,000 researchers finds that science really is "publish or perish."

Science Moneyball: The Secret to a Successful Academic Career

John is a Science contributing correspondent.

(After being alerted to differences between our widget’s prediction and that of the full model, Science consulted with the authors of the Current Biology study and discovered problems in the equations they had provided to us. We have fixed those issues, but stress that this version is a simplified version of their full model and the results will likely differ—this one depends on just six variables, while the original one draws on 50 variables from PubMed records. Still, we hope you will find this version useful as an exploratory tool, for seeing how getting more papers or citations improves your relative likelihood of becoming a PI.)

For biomedical researchers who aspire to run their own labs, becoming so-called principal investigators (PIs), the secret is to publish frequently, as first author, and in top journals. That career advice may seem obvious, but this time it’s backed up by a new analysis of data scraped from PubMed, the massive public repository of biological abstracts.

The study, reported today in Current Biology, uses the status of last author as a proxy for academic success. Those corresponding authors are likely to be running their own labs, the brass ring that young researchers are trying to grab. See what your chances are using Science’s PI Predictor graph.

The tool was dreamt up by David van Dijk, a computational biologist doing a postdoc at the Weizmann Institute of Science in Rehovot, Israel. If “publish or perish” really is the main determinant of academic job success, then he figured that the publication records of scientists living today should reveal that slant. So he teamed up with Ohad Manor and Lucas Carey, computational biologists at the University of Washington, Seattle, and Pompeu Fabra University in Barcelona, Spain, respectively, to look for statistical patterns in career trajectories.

Drawing on 23 million abstracts in PubMed, the team focused on people whose full names are unique in the repository. They were also interested in gender as a hiring factor, so they used a database of first names to classify scientists as men and women, excluding those with ambiguous names. After all that filtering, they ended up with publication records of 25,604 scientists.

Their analysis incorporated more than 200 variables, from the global rank of a scientist’s university, to the total number of papers, citations to those papers, and the impact factor of the journals in which they appeared. One revelation: The first few years of papers are enough to predict who will become a PI. Another is that impact factor isn't everything. At least later in a career, a large number of publications in low-ranking journals can be just as good as a few in the big ones. That’s “perhaps the most interesting finding,” says Sam Gershman, a computational neuroscientist at the Massachusetts Institute of Technology in Cambridge.

The prestige of journals is still too strong of a determinant of career destiny, Gershman argues. “If impact factor eventually becomes less important than the number of citations, our relationship to journal publishers will profoundly change,” he hopes. “We might no longer need publication in prestigious journals to further our careers, as long as our publications are highly cited. For now, however, "impact factor is still the strongest predictor of becoming a PI.”

Carey says that the tool could be expanded to include data about grants funding, tenure, and other career features. “This type of analysis might suggest candidates with a high probability of success that hiring committees might initially overlook,” he says. “I’m thinking of Moneyball and the Oakland A’s,” he says, referring to the book and film about using statistical analysis to build a better baseball team. Carey says that the role of Billy Beane, the general manager who applied the technique, could be played by “a very charismatic website to which hiring committees could submit CVs.”

*Correction, 2 June, 1:20 p.m.: The affiliations for Manor and Carey have been corrected.

*Update, 3 June, 9:20 a.m.: This article has been updated to define a PI (principal investigator).

Posted in Biology, Math, Scientific Community Scientific Publishing