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
listen; there's a hell of a good universe next door: let's go.e.e. cummingsgo theremore quotes

design: fun



UCD Computational and Molecular Biology Symposium, Dublin, Ireland. 1-2 Dec 2016.


art + science

Bloomberg Businessweek Design Conference — San Francisco, 2013

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Design loves science and science loves design, but doesn't always know it. (Bloomberg Businessweek)

science design

Together with Alberto Cairo, I presented at the Bloomberg Businessweek Design Conference (highlights) on the topic of design and communication in the sciences.

Alberto, as the journalist, motivated why communication should include access to detail through an engaging narrative. He made the distinction between the specialist (heavy on detail) and the communicator (focus on narrative) and emphasized that the distinction is artificial, though often played out (watch video).

I, as the scientist, underscored the importance of clear communication between scientists. As the specialists, they are often very poor communicators. Pick up any science journal and you'll quickly discover that scientists either aren't good at telling stories or are discouraged to do so by the medium. The consequence is the same: papers read like a wall of text. TL;DR anyone? The quality of visual communication in general ranges from muddled to abysmal (watch video).

We need more leaders in the field, such as Nature Publishing Group, to reward and emphasize good visual communication (e.g. Nature Cancer Review 2013 Figure Calendar).

Our presentations concluded with a 15 minute moderated discussion with Sam Grobart, senior Businesssweek writer. Everyone got a little cheeky. Good fun.

presentation video

Watch: my presentation, conversation with Alberto Cairo, moderated by Sam Grobart. (Bloomberg TV), Albert Cairo's presentation.

presentation slides

This was a lightning 7 minute talk. I did more planning about what to say than I usually do, given that there was virtually no opportunity for any kind of backtracking, and include a running narrative below each slide.

Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
Martin Krzywinski - Bloomberg Businessweek Design Conference 2013
1/32

download presentation

My slides are available as PDF, keynote (zipped) or Quicktime. The format is 16:9 HD.

Bloomberg Businessweek Design Issue

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
The reality of redesign is disruptive. How can we pursue new ideas and opportunities without leaving consumers confused or angry? Businessweek puts that question to some of the world's most accomplished designers. (Bloomberg Businessweek Design Issue)

On 28 Jan 2013, Bloomberg Businessweek Design Issue will capture the ideas from the conference and the personalities that generated them.

During the conference, each talk was captured in a series of sketches by Tom Wujec: my talk sketch and moderated discussion sketch.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Date completed: ongoing — an accurate assessment of the state of the visual communication field in science. (read article)
VIEW ALL

news + thoughts

Classifier Evaluation

Fri 05-08-2016

It is important to understand both what a classification metric expresses and what it hides.

We examine various metrics use to assess the performance of a classifier. We show that a single metric is insufficient to capture performance—for any metric, a variety of scenarios yield the same value.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Nature Methods Points of Significance column: Classifier Evaluation (read)

We also discuss ROC and AUC curves and how their interpretation changes based on class balance.

Altman, N. & Krzywinski, M. (2016) Points of Significance: Classifier evaluation. Nature Methods 13:603-604.

...more about the Points of Significance column

Happy 2016 `\pi` Approximation, roughly speaking

Sun 24-07-2016

Today is the day and it's hardly an approximation. In fact, `22/7` is 20% more accurate of a representation of `\pi` than `3.14`!

Time to celebrate, graphically. This year I do so with perfect packing of circles that embody the approximation.

By warping the circle by 8% along one axis, we can create a shape whose ratio of circumference to diameter, taken as twice the average radius, is 22/7.

If you prefer something more accurate, check out art from previous `\pi` days: 2013 `\pi` Day and 2014 `\pi` Day, 2015 `\pi` Day, and 2016 `\pi` Day.

Logistic Regression

Wed 13-07-2016

Regression can be used on categorical responses to estimate probabilities and to classify.

The next column in our series on regression deals with how to classify categorical data.

We show how linear regression can be used for classification and demonstrate that it can be unreliable in the presence of outliers. Using a logistic regression, which fits a linear model to the log odds ratio, improves robustness.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Nature Methods Points of Significance column: Logistic regression? (read)

Logistic regression is solved numerically and in most cases, the maximum-likelihood estimates are unique and optimal. However, when the classes are perfectly separable, the numerical approach fails because there is an infinite number of solutions.

Altman, N. & Krzywinski, M. (2016) Points of Significance: Logistic regression. Nature Methods 13:541-542.

Background reading

Altman, N. & Krzywinski, M. (2016) Points of Significance: Regression diagnostics? Nature Methods 13:385-386.

Altman, N. & Krzywinski, M. (2015) Points of Significance: Multiple Linear Regression Nature Methods 12:1103-1104.

Altman, N. & Krzywinski, M. (2015) Points of significance: Simple Linear Regression Nature Methods 12:999-1000.

...more about the Points of Significance column

Visualizing Clonal Evolution in Cancer

Thu 02-06-2016

Genomic instability is one of the defining characteristics of cancer and within a tumor, which is an ever-evolving population of cells, there are many genomes. Mutations accumulate and propagate to create subpopulations and these groups of cells, called clones, may respond differently to treatment.

It is now possible to sequence individual cells within a tumor to create a profile of genomes. This profile changes with time, both in the kinds of mutation that are found and in their proportion in the overall population.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Ways to present temporal and phylogenetic evolution of clones in cancer. M Krzywinski (2016) Molecular Cell 62:652-656. (read)

Clone evolution diagrams visualize these data. These diagrams can be qualitative, showing only trends, or quantitative, showing temporal and population changes to scale. In this Molecular Cell forum article I provide guidelines for drawing these diagrams, focusing with how to use color and navigational elements, such as grids, to clarify the relationships between clones.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
How to draw clone evolution diagrams better. M Krzywinski (2016) Molecular Cell xxx:xxx-xxx. (read)

I'd like to thank Maia Smith and Cydney Nielsen for assistance in preparing some of the figures in the paper.

Krzywinski, M. (2016) Visualizing Clonal Evolution in Cancer. Mol Cell 62:652-656.

Binning High-Resolution Data

Wed 01-06-2016

Limitations in print resolution and visual acuity impose limits on data density and detail.

Your printer can print at 1,200 or 2,400 dots per inch. At reading distance, your reader can resolve about 200–300 lines per inch. This large gap—how finely we can print and how well we can see—can create problems when we don't take visual acuity into account.

Martin Krzywinski @MKrzywinski mkweb.bcgsc.ca
Nature Methods Points of View column: Binning high-resolution data. (read)

The column provides some guidelines—particularly relevant when showing whole-genome data, where the scale of elements of interest such as genes is below the visual acuity limit—for binning data so that they are represented by elements that can be comfortably discerned.

Krzywinski, M. (2016) Points of view: Binning high-resolution data. Nature Methods 13:463.

...more about the Points of View column