Make sure the mutation class isn't affecting the data

[petercombs]

Sur pointed out that if we think that the really high AT content is an issue, then it's possible we are biasing our discovery and quantitation of SNPs towards {A,T}->{G,C}.

Looking at the image, I think it's not exactly transition vs transversion that I care about, but it's at least along the right lines.

I would predict it won't make a huge difference—doing both the stalks and spores should control for this quite well. Nevertheless, if Sur thought of it, then a reviewer likely will too, so it will be useful to have a figure already prepared that I can point to and prove it's not a real effect.

[petercombs]

So when we limit to SNPs that have some shot of being significant (i.e. where we can perform the Fisher exact test in at least 10 samples), we get

SNPs that increase GC 1437
SNPs that increase AT 937
SNPs that don't change AT/GC 150

Note that this does take into account major/minor alleles (though it should be a relatively small effect). So 1400 SNPs where the most common allele brings the GC content closer to biased, regardless of what's in the reference genome.

So that's actually a pretty significant bias. Now the question is what to do about it...