Wednesday, November 6, 2013

A Mitochondrial Manhattan Plot

Manhattan plots have become the standard way to visualize results for genetic association studies, allowing the viewer to instantly see significant results in the rough context of their genomic position.  Manhattan plots are typically shown on a linear X-axis (although the circos package can be used for radial plots), and this is consistent with the linear representation of the genome in online genome browsers.  Many genetic studies often overlook the other genome within all our cells, the mitochondrial genome. This circular molecule has been shown to be associated (albeit inconsistently) with many disease traits, and functional variants from this genome are now included in most genotyping platforms.

Thanks to the clever work of several graduate students in the Vanderbilt Center for Human Genetics Research (most notably Ben Grady), mitochondrial genetic associations can be visualized in the context of key regions of the mitochondrial genome using a radial plot in the R package ggplot2.

To make this plot, download the code and run the script (alternatively open the script in R and run interactively):

On the command line:

git clone
cd solarplot
R CMD BATCH solarplot.R

GitHub: Mitochondrial solar plot code and data


  1. Thanks for this. I'm having a hard time understanding though--are the white concentric circles the -log(p-values)? If so, then what is the Y axis exactly, if anything? And how can you can get -log(p-values) > 0?

    1. The y-axis on the left side - if you trace it over horizontally to the right, where it intersects the apex of the circle is what that circle represents. You'll see that most points are on the zero line (or very close to it, in -log space). After that the first line is 1, 2, etc., as you go outward among the concentric circles.

  2. As a general question, can we include SNPs from X, Y and mitochondrial chromosomes in the GWAS or multivariate SNP analyses. I am using plink for processing the markers and X,Y and mitochondrial chromosomes markers are disregarded in the QC in plink.

    Your suggestion is helpful.


  3. This is really interesting. I have a few questions:

    1) is the radial design done purposefully to provide an at-a-glance distinction from gene-based manhattan plots? it is harder to read than the usual linear axis, but perhaps there is a good reason, e.g., above idea or some other reason(s)?

    2) i am a stats geek but not a gene/mito guy. is it typical to adjust the p-values for multiple comparisons, i.e., are your example data, for example, raw or adjusted p-values? and how does that affect interpretation of the plot? if raw, do you treat every "outlier" with heavy skepticism? i am assuming that you're working in the classical tradition here.


    1. Dwight,

      1. This is really more of an illustration for how to make such a radial plot with ggplot2. Mitochondrial ideograms with gene annotations are usually displayed in a circular format, so a circular manhattan plot might be more familiar to mitochondria geneticists. Personally, I agree with you - it's difficult to make comparisons of the magnitude of the -logP for different parts of the chromosome, at least more difficult than a linear plot IMHO.

      2. Manhattan plots are rarely adjusted for multiple comparisons. But, most folks agree that in GWAS at least, a p-value of 5e-8 is "genome-wide significant" (assuming 1 million independent tests adjusting at p=0.05). I'm not sure what the consensus procedure in mitochondria association testing is for multiple testing correction.

  4. Check my new blog on bioinformatics and life science informatics in general at


Note: Only a member of this blog may post a comment.

Creative Commons License
Getting Genetics Done by Stephen Turner is licensed under a Creative Commons Attribution-NonCommercial 3.0 Unported License.