With some very smart people, I work on problems in data visualization applied to cancer research and genome analysis. Previously I was involved in fingerprint mapping, system administration, computer security, fashion photography, medical imaging and LHC particle physics. My work is guided by a need to rationalize, make things pretty, combine science with art, mince words, find good questions and help make connections between ideas. All while exercising snark.
The color summarizer generates statistical color summaries of images.
The color summarizer also identifies representative colors in the image by using k-means clustering to group colors into clusters. The centers of each cluster are also reported by name, using my large database of named colors.
Below is an example of a detailed color report of an image—an adorable Fiat 126p I found while it was screaming out its color against the fading background of Havana.
The Brewer color palettes are an excellent source for perceptually uniform color palettes. I provide Adobe Swatches for all colors in the Brewer Palettes.
I also provide a short talk to help you understand why these palettes are important.
Color Palettes for Color Blindness
Color blindness is a thing. You should worry about it when you're designing and especially when you're encoding information.
I provide some background on color blindness and give options for choosing 7-, 12- and 15-color palettes that are colorblind safe.
Probably the world's largest list of named colors.
With more than 8,300 colors, even a mantis shrimp would be impressed. You can finally imagine a color you can't even imagine and name it!
The color name list is hooked into the color summarizer's clustering. You can get a list of words, derived from the color names, that describes an image.
A visual survey of the color proportions in flags of 256 countries.
Flags are depicted by concentric rings whose thickness is a function of the amount of that color in the flag.
I make the flag color catalog available, as well as similarity scores based on color proportions for each flag pair, so you can run your own analysis.
We focus on the important distinction between confidence intervals, typically used to express uncertainty of a sampling statistic such as the mean and, prediction and tolerance intervals, used to make statements about the next value to be drawn from the population.
Confidence intervals provide coverage of a single point—the population mean—with the assurance that the probability of non-coverage is some acceptable value (e.g. 0.05). On the other hand, prediction and tolerance intervals both give information about typical values from the population and the percentage of the population expected to be in the interval. For example, a tolerance interval can be configured to tell us what fraction of sampled values (e.g. 95%) will fall into an interval some fraction of the time (e.g. 95%).
Altman, N. & Krzywinski, M. (2018) Points of significance: Predicting with confidence and tolerance Nature Methods 15:843–844.
Krzywinski, M. & Altman, N. (2013) Points of significance: Importance of being uncertain. Nature Methods 10:809–810.
A 4-day introductory course on genome data parsing and visualization using Circos. Prepared for the Bioinformatics and Genome Analysis course in Institut Pasteur Tunis, Tunis, Tunisia.
Data visualization should be informative and, where possible, tasty.
Stefan Reuscher from Bioscience and Biotechnology Center at Nagoya University celebrates a publication with a Circos cake.
The cake shows an overview of a de-novo assembled genome of a wild rice species Oryza longistaminata.
The presence of constraints in experiments, such as sample size restrictions, awkward blocking or disallowed treatment combinations may make using classical designs very difficult or impossible.
Optimal design is a powerful, general purpose alternative for high quality, statistically grounded designs under nonstandard conditions.
We discuss two types of optimal designs (D-optimal and I-optimal) and show how it can be applied to a scenario with sample size and blocking constraints.
Smucker, B., Krzywinski, M. & Altman, N. (2018) Points of significance: Optimal experimental design Nature Methods 15:599–600.
Krzywinski, M., Altman, N. (2014) Points of significance: Two factor designs. Nature Methods 11:1187–1188.
Krzywinski, M. & Altman, N. (2014) Points of significance: Analysis of variance (ANOVA) and blocking. Nature Methods 11:699–700.
Krzywinski, M. & Altman, N. (2014) Points of significance: Designing comparative experiments. Nature Methods 11:597–598.
An illustration of the Tree of Life, showing some of the key branches.
The tree is drawn as a DNA double helix, with bases colored to encode ribosomal RNA genes from various organisms on the tree.
All living things on earth descended from a single organism called LUCA (last universal common ancestor) and inherited LUCA’s genetic code for basic biological functions, such as translating DNA and creating proteins. Constant genetic mutations shuffled and altered this inheritance and added new genetic material—a process that created the diversity of life we see today. The “tree of life” organizes all organisms based on the extent of shuffling and alteration between them. The full tree has millions of branches and every living organism has its own place at one of the leaves in the tree. The simplified tree shown here depicts all three kingdoms of life: bacteria, archaebacteria and eukaryota. For some organisms a grey bar shows when they first appeared in the tree in millions of years (Ma). The double helix winding around the tree encodes highly conserved ribosomal RNA genes from various organisms.
Johnson, H.L. (2018) The Whole Earth Cataloguer, Sactown, Jun/Jul, p. 89