Martin Krzywinski / Genome Sciences Center / mkweb.bcgsc.ca Martin Krzywinski / Genome Sciences Center / mkweb.bcgsc.ca - contact me Martin Krzywinski / Genome Sciences Center / mkweb.bcgsc.ca on Twitter Martin Krzywinski / Genome Sciences Center / mkweb.bcgsc.ca - Lumondo Photography Martin Krzywinski / Genome Sciences Center / mkweb.bcgsc.ca - Pi Art Martin Krzywinski / Genome Sciences Center / mkweb.bcgsc.ca - Hilbertonians - Creatures on the Hilbert Curve
Mad about you, orchestrally.Hooverphonicfeel the vibe, feel the terror, feel the pain

design: beautiful



Bioinformatics and Genome Analysis Course. Izmir International Biomedicine and Genome Institute, Izmir, Turkey. May 2–14, 2016


visualization + design

Getting into Visualization of Large Biological Data Sets

The 20 imperatives of information design

Martin Krzywinski, Inanc Birol, Steven Jones, Marco Marra

Presented at Biovis 2012 (Visweek 2012). Content is drawn from my book chapter Visualization Principles for Scientific Communication (Martin Krzywinski & Jonathan Corum) in the upcoming open access Cambridge Press book Visualizing biological data - a practical guide (Seán I. O'Donoghue, James B. Procter, Kate Patterson, eds.), a survey of best practices and unsolved problems in biological visualization. This book project was conceptualized and initiated at the Vizbi 2011 conference.

If you are interested in guidelines for data encoding and visualization in biology, see our Visualization Principles Vizbi 2012 Tutorial and Nature Methods Points of View column by Bang Wong.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Getting into Visualization of Large Biological Data Sets. M Krzywinski, I Birol, S Jones, M Marra (poster presentation) (PDF)


The 20 imperatives of information design


ENSURE LEGIBILITY AND FOCUS ON THE MESSAGE

Create legible visualizations with a strong message. Make elements large enough to be resolved comfortably. Bin dense data to avoid sacrificing clarity.

Distinguish between exploration and communication.

Use exploratory tools (e.g. genome browsers) to discover patterns and validate hypotheses. Avoid using screenshots from these applications for communication – they are typically too complex and cluttered with navigational elements to be an effective static figure.

Do not exceed resolution of visual acuity.

Our acuity is ~50 cycles/degree or about 1/200 (0.3 pt) at 10 inches. Ensure the reader can comfortably see detail by limiting resolution to no more than 50% of acuity. Where possible, elements that require visual separation should be at least 1 pt part.

Use no more than ~500 scale intervals.

Ensure data elements are at least 1 pt on a two-column Nature figure (6.22 in), 4 pixels on a 1920 horizontal resolution display, or 2 pixels on a typical LCD projector. These restrictions become challenges for large genomes.

Show variation with statistics.

Data on large genomes must be downsampled. Depict variation with min/max plots and consider hiding it when it is within noise levels. Help the reader notice significant outliers.

Do not draw small elements to scale.

Map size of elements onto clearly legible symbols. Legibility and clarity are more important than precise positioning and sizing. Discretize sizes and positions to facilitate making meaningful comparisons.

Aggregate data for focused theme.

A strong visual message has no uncertainty in its interpretation. Focus on a single theme by aggregating unnecessary detail.

Show density maps and outliers.

Establishing context is helpful when emergent patterns in the data provide a useful perspective on the message. When data sets are large, it is difficult to maintain detail in the context layer because the density of points can visually overwhelm the area of interest. In this case, consider showing only the outliers in the data set.

Consider whether showing the full data set is useful.

The reader’s attention can be focused by increasing the salience of interesting patterns. Other complex data sets, such as networks, are shown more effectively when context is carefully edited or even removed.

USE EFFECTIVE VISUAL ENCODINGS TO ORGANIZE INFORMATION.

Match the visual encoding to the hypothesis. Use encodings specific and sensitive to important patterns. Dense annotations should be independent of the core data in distinct visual layers.

Use the simplest encoding.

Choose concise encodings over elaborate ones.

Help the reader judge accurately.

Accuracy and speed in detecting differences in visual forms depends on how information is presented. We judge relative lengths more accurately than areas, particularly when elements are aligned and adjacent. Our judgment of area is poor because we use length as a proxy, which causes us to systematically underestimate.

Use encodings that are robust and comparable.

In addition to being transparent and predictable, visualizations must be robust with respect to the data. Changes in the data set should be reflected by proportionate changes in the visualization. Be wary of force-directed network layouts, which have low spatial autocorrelation. In general, these are neither sensitive nor specific to patterns of interest.

Crop scale to reveal fine structure in data.

Biological data sets are typically high-resolution (changes at base pair level can meaningful), sparse (distances between changes are orders of magnitude greater than the affected areas) and connect distant regions by adjacency relationships (gene fusions and other rearrangements). It is difficult to take these properties into account on a fixed linear scale, the kind used by traditional genome browsers. To mitigate this, crop and order axis segments arbitrarily and apply a scale adjustment to a segment or portion thereof.

Use perceptual palettes.

Selecting perceptually favorable colors is difficult because most software does not support the required color spaces. Brewer palettes exist for the full range of colors to help us make useful choices. Qualitative palettes have no perceived order of importance. Sequential palettes are suitable for heat maps because they have a natural order and the perceived difference between adjacent colors is constant. Twin hue diverging palettes, are useful for two-sided quantitative encodings, such as immunofluorescence and copy number.

Never use hue to encode magnitude.

Hue does not communicate relative change in values because we perceive hue categorically (blue, green, yellow, etc). Changes within one category have less perceptual impact than transitions between categories. For example, variations across the green/yellow boundary are perceived to be larger than variations across the same sized hue interval in other parts of the spectrum.

USE EFFECTIVE DESIGN PRINCIPLES TO EMPHASIZE AND COMMUNICATE PATTERNS.

Well-designed figures illustrate complex concepts and patterns that may be difficult to express concisely in words. Figures that are clear, concise and attractive are effective – they form a strong connection with the reader and communicate with immediacy. These qualities can be achieved with methods of graphic design, which are based on theories of how we perceive, interpret and organize visual information.

Reduce unnecessary variation.

The reader does not know what is important in a figure and will assume that any spatial or color variation is meaningful. The figure’s variation should come solely from data or act to organize information.

Encapsulate details.

Including details not relevant to the core message of the figure can create confusion. Encapsulation should be done to the same level of detail and to the simplest visual form. Duplication in labels should be avoided.

Use consistent alignment. Center on theme.

Establish equivalence using consistent alignment. Awkward callouts can be avoided if elements are logically placed.

Respect natural hierarchies.

When the data set embodies a natural hierarchy, use an encoding that emphasizes it clearly and memorably. The use hierarchy in layout (e.g. tabular form) and encoding can significantly improve a muddled figure.

Palette for color blindness / Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
This 15-color palette provides good discrimination for common color blindness types. Individuals with tritanopia cannot distinguish colors marked with ● and ◥. (hires)

Be aware of the luminance effect.

Color is a useful encoding – the eye can distinguish about 450 levels of gray, 150 hues, and 10-60 levels of saturation, depending on the color – but our ability to perceive differences varies with context. Adjacent tones with different luminance values can interfere with discrimination, in interaction known as the luminance effect.

Be aware of color blindness.

In an audience of 8 men and 8 women, chances are 50% that at least one has some degree of color blindness. Use a palette that is color-blind safe. In the palette below the 15 colors appear as 5-color tone progressions to those with color blindness. Additional encodings can be achieved with symbols or line thickness.

I have designed 15-color palettes for color blindess for each of the three common types of color blindness.

news + thoughts

Unentangling complex plots

Fri 10-07-2015

The Points of Significance column is on vacation this month.

Meanwhile, we're showing you how to manage small multiple plots in the Points of View column Unentangling Complex Plots.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Nature Methods Points of View column: Unentangling complex plots. (download, more about Points of View)

Data in small multiples can vary in range, noise level and trend. Gregor McInerny and myself show you how you can deal with this by cropped and scaling the multiples to a different range to emphasize relative changes while preserving the context of the full data range to show absolute changes.

McInerny, G. & Krzywinski, M. (2015) Points of View: Unentangling complex plots. Nature Methods 12:591.

...more about the Points of View column

Fixing Jurassic World science visualizations

Fri 10-07-2015

The Jurassic World Creation Lab webpage shows you how one might create a dinosaur from a sample of DNA. First extract, sequence, assemble and fill in the gaps in the DNA and then incubate in an egg and wait.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
We can't get dinosaur genomics right, but we can get it less wrong. (a) Corn genome used in Jurassic World Creation Lab website. Image is from the Science publication B73 Maize Genome: Complexity, Diversity, and Dynamics. Photo and composite by Universal Studios and Amblin Entertainment. (b) Random data on 8 chromosomes from chicken genome resized to triceratops genome size (3.2 Gb). Image by Martin Krzywinski. (c) Actual genome data for lizard genome, UCSC anoCar2.0, May 2010. Image by Martin Krzywinski. Triceratops outline in (b,c) from wikipedia.

With enough time, you'll grow your own brand new dinosaur. Or a stalk of corn ... with more teeth.

What went wrong? Let me explain.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Corn World: Teeth on the Cob.

Printing Genomes

Tue 07-07-2015

You've seen bound volumes of printouts of the human reference genome. But what if at the Genome Sciences Center we wanted to print everything we sequence today?

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Curiously, printing is 44 times as expensive as sequencing. (details)

Gene Volume Control

Thu 11-06-2015

I was commissioned by Scientific American to create an information graphic based on Figure 9 in the landmark Nature Integrative analysis of 111 reference human epigenomes paper.

The original figure details the relationships between more than 100 sequenced epigenomes and genetic traits, including disease like Crohn's and Alzheimer's. These relationships were shown as a heatmap in which the epigenome-trait cell depicted the P value associated with tissue-specific H3K4me1 epigenetic modification in regions of the genome associated with the trait.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Figure 9 from Integrative analysis of 111 reference human epigenomes (Nature (2015) 518 317–330). (details)

As much as I distrust network diagrams, in this case this was the right way to show the data. The network was meticulously laid out by hand to draw attention to the layered groups of diseases of traits.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Network diagram redesign of the heatmap for a select set of traits. Only relationships with –log P > 3.9 are displayed. Appears on Graphic Science page in June 2015 issue of Scientific American. (details)

This was my second information graphic for the Graphic Science page. Last year, I illustrated the extent of differences in the gene sequence of humans, Denisovans, chimps and gorillas.

Sampling distributions and the bootstrap

Thu 11-06-2015

The bootstrap is a computational method that simulates new sample from observed data. These simulated samples can be used to determine how estimates from replicate experiments might be distributed and answer questions about precision and bias.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Nature Methods Points of Significance column: Sampling distributions and the bootstrap. (read)

We discuss both parametric and non-parametric bootstrap. In the former, observed data are fit to a model and then new samples are drawn using the model. In the latter, no model assumption is made and simulated samples are drawn with replacement from the observed data.

Kulesa, A., Krzywinski, M., Blainey, P. & Altman, N (2015) Points of Significance: Sampling distributions and the bootstrap Nature Methods 12:477-478.

Background reading

Krzywinski, M. & Altman, N. (2013) Points of Significance: Importance of being uncertain. Nature Methods 10:809-810.

...more about the Points of Significance column