They still seem to be the rage in some fields, and used to be an incredibly common means of showing gene expression patterns in journals that published such things. But heat maps have fallen out of favor it seems.
True, they take up lots of space, they are essentially a colorful table, and often our eyes are not very good at detecting fine scale differences when looking at color gradation. This final point is important because it raises the possibility that shifts or boundaries may either be undetected or possibly worse, falsely observed.
So heat maps have been relegated to the visualization scrap heap, at least in some people's minds, and seemingly expelled from scientific journals. Despite this, they can help visualize large scale patterns in data and I think that's why some fields continue to use them or are even expanding their use.
In any case, this got me thinking! What would make heat maps, at least those we use to explore changes in gene expression patterns, better? We like the ability to use color to see broad patterns of change, but to ensure we're not fooled by the color we might also want to see the raw data. But printing the raw data on the heat map just looks messy and that's one of the big no-no's of data visualization.
Why not instead, make the heat map interactive! Have it initialize as a standard heat map with the color gradient, but then make specific data points available for perusal, say with the mouse pointer.
Thus was born this very simple example of a gene expression heat map using the hugely powerful program d3js . Click the image on the left see it in action! The code is also hosted as a Gist on the bl.ocks.org repository if you want to see it here http://bl.ocks.org/natemiller/84964461d37181cf055c
With such an example, we could visualize the expression data (upregulated: red, downregulated: green), but also by "mousing-over" (watch out for the technical terminology) individual cells we can see their actual data values. Now this is admittedly stupidly simple example, but it captures the idea and I'd love to see journals catch onto such things.
So yes, heat maps take up valuable page space in print journals and so probably no longer deserve inclusion there...but they do represent a valuable means of conveying information. In our group we argued whether the raw expression data should be supplied along with a published article, possible in an on-line supplement and what form that data should take. One possibility is a huge multi-page table loaded with numbers. Another is a series of colored heat maps (sans raw values). In many cases such raw data does deserved to be shared and a place on the web, but why restrict our options to the utilization a passive ("boring") data table or a "data-less" heat map! Its the web! Let's utilize the power of the web to be interactive and combined the two. The interactive gene expression heat map!
Now we just have to get the journals on board, allowing the publication of web-based data in this type of format.
Just a thought.