Hi all,
We have been sequencing (HiSeq) genomes using FFPE source material for a while now. We have a decent handle on what the base biases are, and the general quality problems that might occur but we're lacking in an explanation of the CNV quality.
The copy number results (doesn't matter which tool) have a lot more noise than they do from fresh sources. Specifically, there are a lot more tiny gains and losses, and a much wider distribution of read depth when using FFPE tissues. We find that the noise is not common between samples and the only way to attenuate it is to do a lot of smoothing, which reduces the detection resolution.
Has anyone else experienced the same problems? I'd also be interested to hear if anyone is finding the opposite where deeply sequencing FFPE genomes give CNV results that are as good as they get from fresh sources.
EDIT: Technicals: We usually have a matched tumour/normal comparison where the genomes are sequenced to the same depth (usually 30-100X). We see the noise when either or both the tumour and normal are from FFPE, with extra noise when they are both FFPE. We have tried a number of tools, all giving similar results. For example, HMMCopy, CNAnorm, Control-FreeC, etc. all of which used GC correction. We have looked independently at GC correction and it doesn't help with FFPE specific noise.
godspeed
Can I ask you what tools you used, what is the typical coverage, if you use a match normal (and if so, if you use same library prep) and if the tools you used perform GC correction. These are the first relevant questions that pop in my mind...
Hi Richard, I am running into similar issues right now and wonder how you dealt with this problem? Do you have any insights to share?