“Interactivity in scientific figures is a key tool for data exploration and the scientific process”

This post was originally published on this site

Last summer we launched our interactive figures initiative with plotly. Since then, we have published 22 interactives figures in seven articles across two platforms (F1000Research and MNI Open Research). Our collaboration with ploty (and Code Ocean) was also covered in a recent Nature Toolbox article.

Authors have taken advantage of the wide range of interactive possibilities available – there are dynamic models, dropdowns that replot multiple datasets, sliders that redraw confidence intervals on multiple graphs simultaneously and cross-linked principle component analyses. We promised to feature the ones we published in early 2018 and here they are. In this post authors describe their figures and share why they wanted to make them interactive.


Peter Rogan: Accurate cytogenetic biodosimetry through automated dicentric chromosome curation and metaphase cell selection

“The advantages are obvious after one plays with interactive figures. If properly employed these are more than a space saving measure; they enable greater insight. Certain parts of our results were made possible by those insights.

In our case, we were testing methods to identify false positives in automated image analysis of chromosomes exposed to radiation. We designed Figure 6 so we could replot data for 16 different samples, and it became apparent that ordering images by combined Z scores was more consistent at discarding false positives than group bin distance; this was true across doses and laboratories, suggesting the Z score approach is robust. The reader can see this too. Previously, we only plotted a single dose and lab, so readers would have to take our word for it. It’s a form of reproducibility.

In our opinion, if interactivity improves understanding of the work that can be reflected in the manuscript text, it is worth adding.”


Benjamin Delory: archiDART v3.0: A new data analysis pipeline allowing the topological analysis of plant root systems

“Our animation shows how a barcode produced by a persistent homology analysis is able to capture the topology of a plant root system. We chose to create an animation instead of a static figure for two reasons: (1) it was the best way to visualize the progression of a mathematical function (here, the geodesic distance) along a root system, and (2) we wanted to demonstrate how a root system and its corresponding barcode are related to each other (birth and death of bars in the persistence barcode). Next to this interactive figure generated with plotly, we also developed a shiny web application to demonstrate the capabilities of the R package presented in our F1000Research article. This app contains several interactive panels where users are allowed to test different scenarios by changing input variables and/or graphical parameters.”


Sara Hagg: Positive bias for European men in peer reviewed applications for faculty position at Karolinska Institutet

“Our F1000Research article argues that there is a bias towards awarding faculty positions to European men, at least at the Karolinska Institutet. In Figure 3, readers can hover over a data point showing the individual application score, and click on a data point on any of the three PCA score plots to highlight individuals that share that characteristic both within and across the PCA plots. For example, clicking on a ‘woman’ data point highlights all women within the Gender PCA as well as the same women in the Ethnicity and Research Area plots. This way readers can easily visualise and compare variance for several traits simultaneously for a subset of individuals. Double clicking the plot resets the figure.”


Chris Hartgerink: Analyzing DECREASE trials to estimate evidence of data manipulation

“Our interactive figure plots the estimated percentage and proportion of data points manipulated in DECREASE-I and –IV, two controversial trials investigating the effect of beta blockers on mortality in non-cardiac surgery. Readers can shift the confidence intervals (CIs) on all four panels simultaneously using an interactive slider. This interactivity is a big help with interpreting the graph, making it easier to see what proportion/number of data points are in the CI, and easily compare these values between the two trials.”


Steven Xijin Ge: PPInfer: a Bioconductor package for inferring functionally related proteins using protein interaction networks

“Like the photos posted on online dating websites, a conventional figure is a carefully chosen snapshot that can be supplemented by interactivity. Interactive figures enable readers to gain a deeper understanding of a dataset by rotating, clicking, zooming, or even rendering from different perspectives. It is about time for interactive plots to transform scientific publishing as (1) scientific datasets are becoming increasingly large and complex, (2) emerging tools like Plotly made it straightforward to add interactivity, and (3) more researchers are reading papers online, instead of on paper. This is why we made some of the figures in our article interactive. My group has also been implementing interactivity in bioinformatics tools such as iDEP for analyzing RNA-Seq data.”


Jonathan Ronen: netSmooth: Network-smoothing based imputation for single cell RNA-seq

“Interactivity in scientific figures is a key tool for data exploration and the scientific process. Modern biology research involves large data sets and many insights may be obtained from exploring the data visually. The interactive figures in the netSmooth paper, the scatter plots in particular, allow the reader to explore the data visually and answer more questions about the results than any single static figure would be able to provide.”


Sebastian Urchs. MIST: A multi-resolution parcellation of functional brain networks

“All the boxplots and scatterplots in our MNI Open Research article on parcelling functional brain networks are interactive, and we created an interactive dashboard so readers can fully explore our parcellations in all their complexity, without the limitations inherent to traditional, static figures. We believe such advances in visualization are important for the open sharing of research outcomes in an era where neuroscience is experiencing an explosion in big data. Interactive visualizations are not just a gimmick, but an important step towards sharing research outcomes without sacrificing their complexity.”

The post “Interactivity in scientific figures is a key tool for data exploration and the scientific process” appeared first on F1000 Blogs.

Comments are closed.

Proudly powered by WordPress | Theme: Baskerville 2 by Anders Noren.

Up ↑