Dealing with abnormally high coverage regions

Question

[novice]

Hello,

I'm seeking methodology advice on a little project I'm working on. I've identified structural variations over an entire genome and found few spots especially rich in variations. I then checked the read coverage over these spots and found a few of them to have an abnormally high coverage (~3X average). I'm not sure how to interpret this correlation. Does this mean the high number of variations in these regions is artificial? How can I test this further? Could I disregard variations in these high-coverage regions for my later analyses?

[Edit] Additional Information:

• Working with S. cerevisiae
• Data is WGS, paired-end
• Used PEM + SP to detect variations
• Verified variations with de-novo assembly

[Floris Brenk]

What kind of data do you have? WGS? exome? RNAseq?

[novice]

Whole Genome Sequencing

[WouterDeCoster]

Try to be complete in your initial posts since something like this is a very important piece of information.

[WouterDeCoster]

How did you identify those spots?

[novice]

Paired-end mapping, split-read mapping, and refinement with de-novo assembly

[WouterDeCoster]

Did you check for low mappability or presence of segmental duplications? Was mapping quality taken into account?

[novice]

Did you check for low mappability or presence of segmental duplications?

No.

Was mapping quality taken into account?

Yes, I used a minimum mapping quality of 20.

[Devon Ryan]

If you happen to be working with human/mouse: do these sites happen to overlap blacklisted regions?

[novice]

I'm working with yeast (S. Cerevisiae). I don't think there's a blacklist available for this species. Sorry for the lack of clarity in the original post!

[pld]

Blacklisted regions?

[Istvan Albert]

I never heard of this concept before but searching for it I found a definition

https://sites.google.com/site/anshulkundaje/projects/blacklists

it says

artifact regions that tend to show artificially high signal

what seems to be lacking is a reasoning of why that be so. I find it a bit excessive to flat out just remove whole regions based on "blaclists". Do people actually do this? Surprised that's all.

[Devon Ryan]

It's standard in ChIPseq, but pretty much no where else (it's normally not useful elsewhere). Some of these seem to be rRNAs or other similar "improperly assembled" regions, which I imagine could cause issues in OPs use-case.