Non-C methylation in Bismark .cov files

[bedward1]

Hi
I am analyzing RRBS using TrimGalore/Bismark and the methylation extractor, and am noticing quite a few non-C (T, A, and G) methylation signals (maybe 20% compared to C, by eye). Is this a normal technical artifact that has been observed?
Thanks

[dpryan]

In the case of G, that is a C, just on the other strand. In the case of T or A, that's either an off by 1 error or the sequence you're looking at is different from the sequence against which the aligning was done.

[bedward1]

Thanks for the replay Devon.
Is the 'off by 1 error' common? What would cause this?

[dpryan]

Off by 1 errors are quite common (common enough that it has its own wikipedia article). The likely cause is the file using 1-based coordinates while whatever you're using to visualize things assumes it's using 0-based coordinates. I remember this being an issue with some of bismark's output at one point, though I don't know what the status of that is. If you look at those and there's an appropriate C (or G, depending on how things were done) just to the right, then that's what's going on.

[fkrueger]

Wouldn't off-by-1 errors more result in 50/50 values than 20%? I'd be happy to have a closer look at whatever problems you are running into, just drop me a line.

[dpryan]

Given the non-random distribution of bases in the genome, probably not.