blog

Björn Brembs is a leading practitioner of Open Science, with a history of not only publishing openly, but also freely sharing the highly specialized research software generated by his lab. Björn’s collaborator, Julien Colomb began developing code to allow the lab to publish their data concomitantly with the analyses.  Taking the sharing of data and code one step further, Björn and Julien have now published their research on founder effects on locomotor behavior of Drosophilia with a dynamic figure that actively runs the code as the reader alters the parameters.

Additionally the article includes an open call to Drosophila researchers to contribute data and help generate the first living scientific figure in a future version of this paper.

To mark the paper passing peer review in just 2 weeks after publication, we caught up with Björn and Julien to talk about the paper and discuss their future plans for this project.

 

F1000Research: Why was this project more suited for publication as a dynamic figure, rather than a static figure?

BB: From the figures in the paper, this was the one where we could test the concept most easily. But in principle, all you need is access to the raw data, the code to evaluate the data and the functionality of the paper to call the data using the software to compute the evaluation. Ideally, rather than having just two different views of the same data as in our case, any reader would eventually be able to open a context menu with different display options, parameter settings and, of course, including the option to see code and data directly to, perhaps, tweak the code in some we the authors didn’t think of.

JC: The short answer is that it is not; any figure produced from tabular data would be suited. What motivated me to incorporate this feature is my belief that the future publications will follow this lead. Let’s imagine what this future could look like: most figures are produced by the publisher and not the scientist: authors have stopped wasting time on figure aesthetics, and readers have all the data represented in their favourite plot. The data is directly accessible for further analysis. In addition, each journal has its own aesthetics, making it easy to spot the publisher of a figure you see in a PowerPoint presentation.

 

F1000Research: Can you briefly outline the technical details behind the dynamic figure, for example what is the scripting language and environment?

JC: We delivered the data and the R code producing the two versions of the figure: we used the ggplot2 package to produce each .png files. An interesting feature of this package is that plots are drawn using specific themes that control its aesthetics. One can for instance produce a grey figure (for printing) changing only one line of code.

The F1000Research team did the rest and incorporated it using Rstudio on their linux server.

 

F1000Research: What are your aims for the upcoming living figure? How can other labs contribute to this work?

BB: We don’t know how many sub-strains of nominally identical stocks differ from the published ‘standard’. The only way to find out is to measure different strains from around the world. Rather than doing all of the testing ourselves, we thought it would be interesting to see what other labs observe themselves. Obviously, that carries the risk that different fly food recipes, breeding conditions and other uncontrollable factors distort the results, but at least in our hands the results were quite robust, so we felt it was worth a try. After all, these tests are supposed to be reproducible among different laboratories! If they are not, the results might point us to ways to improve the tests and make them more reproducible.

JC: Indeed, by having people testing the same sub-strains as we did, we will be able to quantify if the variability seen between different labs is due to differences in lab conditions or differences in sub-strains genetics.

BB: We’d like to see if the results from different labs also fall into the groups we observed or if the data become more of an amorphous cloud spanning the state space of the flies.

Contributing is easy: for now, all we need is coordinate data from 5 minutes of experiment in Buridan’s paradigm as described in the methods section of our paper. In a future version of the figure, there will be a place to enter some technical details and then to upload the data such that the figure will be re-calculated as soon as the data is received. Colleagues should also be able to subscribe to the figure, in order to be alerted when new data has been added.

JC: We are working with F1000Research to make the figure fully functional. In a later phase, we would like to incorporate a similar function directly into the open source trajectory analysis software, such that the data will be uploaded to a server by default, being either openly accessible or ready to become open on publication, depending on the scientists’ wish.

 

F1000Research: Based on your experience, do you have any advice for others who might wish to incorporate a dynamic figure into their own peer-reviewed article?

BB: Make sure your code is clean, understandable and as thoroughly debugged as possible!

JC: To get a dynamic figure, you first need to have the data at hand, in a format that allows an analysis with R. This will be best obtained if you manage your data archiving before you actually start to produce any data. By organizing your data (i.e. rather than using multiple folders, get all your metadata in a unique file which links to the raw data; or have all your data in a single file if you can), you will make it easy to find it using computer commands. Using R to produce figures means that you can export the data with little hassle (you can even publish it directly with R).

Once you have the data, you only need to produce different representation of the data using R (or modify our published code to fit your data), F1000Research will do the rest.

 

F1000Research: Why did you choose to publish this article with us?

BB: The data plotter demoed by F1000Research at a recent conference was already quite close to what we needed, so I suggested the project and F1000Research was quite amenable to our suggestions. If publishers in general were so approachable and flexible, our literature wouldn’t be the dysfunctional mess it is now.

JC: An additional motivation was to test post-publication peer review.  With more than 1500 views before receiving the first peer review report, I think the test was a success. This paper was particularly well suited for this way of publishing, because it highlights a problem the whole fly community has to deal with: differences in the genome of similarly called CS sub-strains.

 

If you have a project that would benefit from being published with dynamic figures, email us at research@f1000.com.