Scientists pour countless hours into their research, meticulously gathering data and uncovering new insights. But effectively communicating these findings to the broader scientific community requires a different skill set: visual storytelling. That’s where well-designed figures come in.

We sat down with Patrice Salomé of Plant Editors, where he works as a scientific editor for text and figures, to learn more about the art and science of creating effective and aesthetically pleasing figures for research manuscripts. Read on to discover some common pitfalls, best practices, and how figure editing services by Plant Editors can help your research shine.

Q: Tell us a little bit about yourself and why authors should listen to you on how to generate “good” figures?

A: I wanted to be a geologist as a kid (I loved shiny crystals and the intricate patterns of agate slices); when I realized how much physics I would need to learn (I’ve always hated physics class), I thought that perhaps I could be an architect. And when I understood how important physics was to designing buildings, I decided that plant biology would be my best option, having already invaded my parents’ living room with house plants.

Fast-forward to graduate school, where I used Arabidopsis to learn about how photoreceptors and temperature keep the circadian clock ticking on time. The lab was using Cricket Graph (this was a long time ago; this graphics software is now called Kaleidagraph), which had (and Kaleidagraph still has) this “template” option: make one graph and set up axis length and style once; you can then use the same graph as template for all related plots, keeping the same plot size, with no adjustments!

This concept of following the same format for related plots was truly foundational in my assembling research figures. With over two decades of practice, I generate figures that look great AND are informative.

Q: What’s one common challenge you see authors face when creating figures?

A: There are two main types of figures: data figures and model figures, the latter being used to summarize the information of the article into a version of a graphical abstract.

Focusing on model figures for a moment, I find that authors are very focused on the specific new result they’ve discovered: Protein X regulates the expression of gene Y, for example. This type of information will generally be shown as two boxes with a connecting line between them. 

What is lacking is context: in which tissue does this regulation take place? Does the localization of protein X change? Was this information obtained by analyzing a mutant in X and/or Y? While that focus is understandable, the broader context can sometimes be lost. My goal is to help scientists take a big step back and create a model figure that effectively positions new discoveries within the larger cellular or tissue environment where these relationships occur. It’s about showing not just the “what,” but also the “where” and “how”. Showing two conditions can be tremendously effective in explaining the new findings: Under normal growth conditions and under stress, for instance; or showing what happens in a wild-type plant and in the mutant of X or Y. In this case, more is better.

For data figures, a major issue is the size and placement of all elements and text, which is just one aspect of accessibility.

Q: Indeed, so how does accessibility factor into assembling scientific figures?

A: Accessibility is absolutely critical. A key consideration is using colorblind-friendly color palettes to ensure that the information conveyed by different colors will be perceived by everyone. I avoid green/red and green/orange combinations, as these two colors will look identical to colorblind readers (and reviewers). This point applies to data figures and model figures

Beyond color, I recommend conveying the information encoded by these colors with different symbols. This way, if color isn’t perfectly perceived, the message still gets across.

Another aspect of accessibility that often gets overlooked is font size, which is directly related to plot size and file size. Datasets can be incredibly large and packed with dense information, requiring the assembly of figures with multiple plots to show all appropriate controls and test samples.

Authors sometimes place all these data panels on one page, but then adjust the final size of the document to make everything fit. It is not uncommon to receive figures of over 50 cm in width, when a printed page will be at most 18 cm for a double-column figure. If font size in each panel is 8 points, the final effective font size will be less than 3 for a 50-cm figure. Too small!I recommend a minimum font size of 6 points, but 8 points is typically best for optimal readability. Here is a super useful trick to avoid issues with font size: set the dimensions of your figure BEFORE populating it with plots and images. Make everything fit into the allocated space, rather than expanding canvas size.

Q: Let’s talk about the visual elements of a plot itself. Any common design advice you offer?

A: Absolutely. I try to make sure that the axes of a plot never draw more attention than the data they illustrate. The axes are there to frame the data and provide scales, not to be the main visual focus.Also, when you’re bringing multiple panels together to form one larger figure, making sure that the individual panels are properly aligned helps create a clean, professional, and easy-to-follow visual narrative. Have you ever been distracted by panels placed somewhat haphazardly on the page? I have!

Q: Are there ever times when you suggest a scientist rethink how they’re plotting their data?

A: Definitely. The type of plot an author initially chooses might not most effectively showcase their data. For instance, while bar graphs showing means and standard deviations are common, a violin plot or a boxplot can often provide a much richer and more informative representation of the data distribution. I’ll sometimes recommend these alternatives to authors.

While I am on the topic of means and standard deviations: There are specific cases when it is appropriate to use them in a graph. Showing a standard deviation is only possible when your data follow a normal distribution. You cannot tell with only two data points (n = 2), so one should instead plot the individual data points, with no mean and no error bar. In fact, many journals ask authors to show all individual data points for n <6 (or n <30 for others). I think it is their way of making sure that you know about the proper use of means and standard deviations without asking explicitly.And another thing, this time about heatmaps and color scales. There are two basic types of color scales for heatmaps: sequential and diverging. A sequential scale is best when showing raw values like FPKM values or counts, with values starting at zero. A sequential scale goes from white to a gradually darker shade of a color. A diverging scale is best for normalized data centered on zero, like Log₂(fold-change) or Z-score-normalized data. In this case, negative values are shown in a gradually darker shade of one color (let’s say blue) for more negative values, while positive values are shown in a gradually darker shade of another color (let’s say red). Again, zero is shown in white. Using the wrong color scale misleads the reader when they look at a heatmap, so everyone should be careful when plotting heatmaps!

Q: File formats can be a bit of a headache. What’s your recommendation for saving plots for manuscript submission?

A: This is a big one! I always recommend saving your plots as PDF files. The beauty of PDFs (a vector file format) is that they remain fully editable when they’re assembled into the final figure. This is a major time-saver when you need to adjust font size and line thickness across multiple panels. TIFF files, by contrast, are in raster format, not vector format, and are therefore not editable. If text isn’t in the right font type or size in a TIFF file, you need to cover it up and write over it, which is rather inefficient and might lead to issues during production of the formatted PDF file of your accepted article.

The only caveat with PDFs is that they are by and large not editable when imported into Microsoft PowerPoint. When and if at all possible, I’d recommend avoiding PowerPoint for final figure assembly.

Q: What about fonts? Any strong opinions there for scientific figures?

A: Oh, have I got opinions on fonts! I find that Arial and Helvetica consistently give the most professional and clean look to a scientific figure, as you can see from the sample model above. Many journals also ask for a sans-serif font, such as Arial and Helvetica. They’re clear, modern, and easy to read. I would stay away from Times New Roman; it looks outdated in a scientific context, at least to me.And one more thing about fonts: I advise against using bold type font in figures. I find that it makes the figures look dark and less inviting than text in a regular font. I suspect that the use of bold type dates back to the days of preparing figures for talks given on old slide decks (I am old enough to remember those). The goal then was to ensure that the attendees in the back of the room could see the text clearly. No one really uses slide decks anymore, so it’s time to ditch the bold!

Ready to make your research figures truly stand out? If you’re looking to elevate the visual impact and clarity of your manuscript figures, learn more about Plant Editors figure services in our blog post on Figure Review Services. Don’t miss our Image Integrity Scans too. And if you’re ready to get a free quote for figure services, contact us on our Quote Page.