Thursday, December 13, 2012

Blind to colorblindness

Since venturing into the more biology-oriented side of organic chemistry, I've noticed something in chemical biology and biochemistry papers that really isn't a problem in synthetic journals. Color. Biologists/chemical biologists love using color. And that makes sense; the meaning of a figure can be more easily conveyed by dyanamic coloration. Consider the following micrograph from an ACS Chemical Biology article (DOI: 10.1021/cb300171p). The figure depicts measurement of zinc levels in mitochondria in different cell types. It's a nice looking figure, and reasonably clear.


But then you go to print the articles (some of us do that) on your black and white printer and the figures become incomprehensible.



Of course, the obvious question: so what? Why not use a color printer? Or simply view things on the computer or a tablet? Content is delivered overwhelmingly by the web now; isn't it natural that figures reflect the most common viewing medium?

Well, yes. But it's still a problem for a sizeable minority (not many women, but as many as 10% of men) of the population who experience colorblindness. "Colorblindness" is a blanket term to refer to any disorder resulting in loss of color perception. Completely grayscale colorblindness is very rare, but difficulty telling red and green apart is relatively common. Variability is seen in the degree of impairment; many people are simply "red-green weak", whereas others cannot tell the two apart.  Blue and yellow are also indistinguishable to some people, though this is less common.

The previously mentioned figure would look like this to someone with protanopia, a form of red-green colorblindness (note that colorblind images were simulated using Photoshop and the Vischeck plugin).


An important thing to note is that what appears illegible in grayscale is not necessarily illegible for colorblindness; a comparative example is shown below  Thus, it's possible to use colors effectively and still accommodate the colorblind population. Take this example from a Merck drug discovery paper (DOI: 10.1021/cb2003225). The quality of the original image (left) is not significantly damaged for a person with red-green colorblindness (center), even though the figure is indiscernable when rendered grayscale (right).


Web developers take colorblindness into account. It's an important topic in website design; professional web designers must ensure their content renders to readers who can't see, can't hear, can't see well, use Apple products, or other disabilities. As such, guidelines have been set up to ensure accessibility, including recommendations for color and contrast. The W3C standards, for example, "help make the Web accessible to people with disabilities including auditory, cognitive, neurological, physical, speech, and visual disabilities" (read more here if you're interested). Validation tools are commonly employed to ensure websites meet these standards.

So why haven't some journals gotten the message?

Consider the following examples, shown in the original form on the top/left with simulated protanopia beneath/right. Both of the example figures are from a 2008 JACS article (DOI: 10.1021/ja807872s). In this first example,  the graph is illustrating the fluorescence response of a bisboronic acid rhodamine derivative to various peptide substrates in order to demonstrate its selectivity. First off, this figure is difficult to read in the first place. But in the original full color image, peptides 1 and 10 are clearly the top symbols. In the colorblind-simulation image, it's tougher to tell. Is it 10 or 6 or 4? Is it 1 or 7? 9 kind of looks close, too. There's the little lines behind the markers, but those are confusing themselves. While the text clarifies this figure further, the figure label does not.

The next example is even worse. The results of labeling studies are shown. In their original color (left) they're quite striking. But the labels are red and green. Hence, under red-green colorblindness, the pictures are meaningless.


This should be unacceptable; all modern fluorescence imaging programs have the option of changing the display color of a particular fluorescent channel; hence, it is trivial to simply recolor the red or green channel to another color (i.e. blue). Take the following figure from a Nature Methods article (DOI: 10.1038/nmeth735) by Alice Ting. Here, the authors have false-colored the fluorescent dyes (Alexa-568 is in reality red and has been colored orange; CFP is cyan-fluorescent protein and is usually cyan anyway but has been adjusted). As shown from comparing the original figure (left) and the colorblind-simulated image (right), very little detail is lost and the figure is clear.


What do the publishers say about color accessibility? Not much. Consider ACS's policy from its Author Guidelines for JACS:
Color. The use of color to enhance the clarity of complex structures, figures, spectra, schemes, etc. is encouraged. Color reproduction of graphics will be provided at no cost to the author. Color graphics should be submitted in CYMK, not RGB, color mode. Graphics intended to appear in black and white or grayscale should not be submitted in color.
Okay. No mention of accessibility. And JACS is fairly broad, and most of its TOC images even contain color. Let's also briefly consider the color policies of a few other journals: ACS Chemical Biology, Cell, Nature Chemistry, Science, PNAS, and PLoS One. There's not much point in reproducing the individual journal guidelines here, but ACS Chem. Bio., Cell, Nature Chemistry, PNAS, and PLos One all say essentially the same thing as JACS, with no mention of handicap accessibility.

The figure preparation guidelines for Science condense colorblindness to one concise sentence:
Avoid using combinations of red and green together.
Some journals are aware; the author guidelines for Nature Chemical Biology, for instance, make some recommendations (interestingly, Nature's formatting guide doesn't mention them, nor does Nature Chemistry):
Authors are encouraged to consider the needs of colorblind readers (a substantial minority of the male population) when choosing colors for figures. Many colorblind readers cannot interpret visuals that rely on discrimination of green and red, for example. Thus, we ask authors to recolor green-and-red heatmaps, graphs and schematics for which colors are chosen arbitrarily. Recoloring primary data, such as fluorescence or rainbow pseudo-colored images, to color-safe combinations such as green and magenta, turquoise and red, yellow and blue or other accessible color palettes is strongly encouraged.
Granted, for some applications, color is unavoidable; take fluorescent microscopy for example. But the ease of making color figures/graphs/charts/schemes has outpaced some journals' awareness of accessibility standards.

In practice, I think the Nature publishing group does a pretty good job with this. And it's possible these things are perhaps checked in the review process even if not mentioned in the author submission guidelines. But I don't think it would hurt to provide this information upfront.

3 comments:

  1. Apart from colorblindness, there's also the thing with multicolor bands ("rainbows") that we learned at an early age to segment colors into color-names. And each of us does that differently.

    We are much less likely to mentally segment luminance bands (single color, dark to white) into bogus abstract names. So people should use that.

    ReplyDelete
  2. I agree with you on that, and I think using color gradients sets you up to easily make trends more/less apparent than they really are.

    You may find this interesting (I was alerted to it via Twitter):

    http://www.dentistry.bham.ac.uk/medicalimaging/page2.asp

    ReplyDelete