This video series focuses on relevant and practical concepts in scientific data visualization. My aim is to make you think more clearly about visual presentation and to make you a better communicator. Each video is about 15 minutes long and comes with a slide deck of the images used in the video, exercise and suggested solutions.
Each video in the series presents fundamental ideas and is designed to provide constraints and guidance to your thoughts about communicating your data. The purpose of scientific data visualization is not merely to inform but also to answer and generate hypotheses.
Whatever your communication medium, you should always have consistency (good!), redundancy (bad!) and an appropriate mapping between relevant and salience in mind (tricky!). Once these are satisfied, look to flow and density of material to achieve clarity (elusive!).
I present these essential topics using biological data as examples. But if you're not a biologist, don't worry. Instead, think about the data structure rather than meaning and you'll be fine.
Download all course materials.
Make it easy to answer relevant questions.
When you think of data visualization, the first ideas that come to mind are a scatter plot, or a bar char, a box plot or a network diagram. These are all data encodings—methods that relate data values to the positions, sizes and shapes of the lines or symbols that appear on the screen or in a figure. There are many data encodings—which do you choose?
watch | PDF slides
Intuitively encode role and relevance.
Shapes and glyphs are really important. They make up the heart of a lot of data plots. Your default should be the circle. If you need different shapes, try to map the classes as intuitively as possible onto the shapes. Use less prominent symbols for data that are less relevant (such as reference data included for context).
watch | PDF slides
Use it for emphasis and visual separation.
Color is one of the most exciting ways in which you can completely screw over your visualization. What can start off as a great diagram can be absolutely ruined by a lack of color judgment. When using color, ask yourself—do I need it? Try to work around it using grey tones from Brewer palettes. If you succeed, you’re in a perfect place to use spot color, sparingly, for emphasis.
watch | PDF slides
Don't make errors in error bars.
Knowing the limits of your knowledge is very important. In biology, it’s important to be able to sample the extent of biological variation. And so being able to show this and other forms of variation in measurements or any computed values in visualizations is very important—it addresses reproducibility and your capacity to make statistical inference. Often this is done with error bars. Ironically, there’s a lot of error associated with the use of and interpretation of error bars.
watch | PDF slides
Organize and clarify.
Design plays a large role in data visualization. Think of design as choreography for the page. In our context it’s not merely driven by aesthetic, but function. Although there’s always room for aesthetic—gently applied—and I really encourage you to find your own and continue to refine it. But always remember, be understood before being articulate. Be legible before being attractive! Your goal here isn’t to make inroads on the global stage of aesthetic studies. Become a good visual explainer. It’s harder … and more worth doing.
watch | PDF slides
No data, no ink.
Data-to-ink ratio, taken to the extreme: if there is no data to show, no ink should be used. The idea of “no data to show” may correspond to a variety of scenarios. There may be sincerely no data to show—no values were collected. Or, there are no significant changes to see. Where possible, you should use empty space to indicate lack of data or lack of change in data. You should never be distracted by something that isn’t relevant and empty space is not distracting—it really just provides contrast to adjacent elements, which presumably correspond to actual data or actionable data.
watch | PDF slides
Respect type and use it to establish hierarchy.
Open up a journal or your favourite text book. Find a figure. There’s probably some labels in there. Maybe it’s a multi-panel figure and the labels are the titles. Maybe there are some callouts that tell you what the parts are. If it’s a plot there are probably axis labels and tick labels and maybe a legend with some labels. There’s usually several informational layers in the image, each with their own labels. These labels should reflect that these layers are different. They should also reflect the relative importance of these layers.
watch | PDF slides
Creating a visualization for Scientific American Graphic Science: from start to finish.
Let’s now look at the process of designing a visualization from scratch—from the encoding all the way to design. This was a graphic I did for the June 2015 issue of Scientific American. It appeared on the Graphic Science page.
watch | PDF slides
Nature uses only the longest threads to weave her patterns, so that each small piece of her fabric reveals the organization of the entire tapestry. – Richard Feynman
Following up on our Neural network primer column, this month we explore a different kind of network architecture: a convolutional network.
The convolutional network replaces the hidden layer of a fully connected network (FCN) with one or more filters (a kind of neuron that looks at the input within a narrow window).
Even through convolutional networks have far fewer neurons that an FCN, they can perform substantially better for certain kinds of problems, such as sequence motif detection.
Derry, A., Krzywinski, M & Altman, N. (2023) Points of significance: Convolutional neural networks. Nature Methods 20:.
Derry, A., Krzywinski, M. & Altman, N. (2023) Points of significance: Neural network primer. Nature Methods 20:165–167.
Lever, J., Krzywinski, M. & Altman, N. (2016) Points of significance: Logistic regression. Nature Methods 13:541–542.
Nature is often hidden, sometimes overcome, seldom extinguished. —Francis Bacon
In the first of a series of columns about neural networks, we introduce them with an intuitive approach that draws from our discussion about logistic regression.
Simple neural networks are just a chain of linear regressions. And, although neural network models can get very complicated, their essence can be understood in terms of relatively basic principles.
We show how neural network components (neurons) can be arranged in the network and discuss the ideas of hidden layers. Using a simple data set we show how even a 3-neuron neural network can already model relatively complicated data patterns.
Derry, A., Krzywinski, M & Altman, N. (2023) Points of significance: Neural network primer. Nature Methods 20:165–167.
Lever, J., Krzywinski, M. & Altman, N. (2016) Points of significance: Logistic regression. Nature Methods 13:541–542.
Our cover on the 11 January 2023 Cell Genomics issue depicts the process of determining the parent-of-origin using differential methylation of alleles at imprinted regions (iDMRs) is imagined as a circuit.
Designed in collaboration with with Carlos Urzua.
Akbari, V. et al. Parent-of-origin detection and chromosome-scale haplotyping using long-read DNA methylation sequencing and Strand-seq (2023) Cell Genomics 3(1).
Browse my gallery of cover designs.
My cover design on the 6 January 2023 Science Advances issue depicts DNA sequencing read translation in high-dimensional space. The image showss 672 bases of sequencing barcodes generated by three different single-cell RNA sequencing platforms were encoded as oriented triangles on the faces of three 7-dimensional cubes.
More details about the design.
Kijima, Y. et al. A universal sequencing read interpreter (2023) Science Advances 9.
Browse my gallery of cover designs.
If you sit on the sofa for your entire life, you’re running a higher risk of getting heart disease and cancer. —Alex Honnold, American rock climber
In a follow-up to our Survival analysis — time-to-event data and censoring article, we look at how regression can be used to account for additional risk factors in survival analysis.
We explore accelerated failure time regression (AFTR) and the Cox Proportional Hazards model (Cox PH).
Dey, T., Lipsitz, S.R., Cooper, Z., Trinh, Q., Krzywinski, M & Altman, N. (2022) Points of significance: Regression modeling of time-to-event data with censoring. Nature Methods 19:1513–1515.