Noisy CNV background in old FFPE sample

Question

[CY]

This may not be a pure bioinformatics question although I got this problem while trying detecting CNV on a old (over 3 years) FFPE samples. What I observed is very noisy CNV backgroud almost acorss all these old FFPE samples.

I did some research and articles shows that this is typical to old FFPE sample are is caused by de-crosslinking during DNA extraction (Pauline Robbe and Haile S). However, I did not get the logics why de-crosslinging on the old FFPE sample could cause such noisy CNV background. Can anyone share some insights? Really appreciated.

Answer 1

[Eric T.]

The samples' DNA may have degraded to the point where you're reading relatively few unique sequence fragments in your sequencing runs of these samples, so regardless of PCR amplification, the deduplicated coverage depth of the aligned reads is lower than you need for accurate CNV estimation.

Depending on how you're doing CNV detection, you may be able to use larger bins for aggregating read counts/depth, trading resolution in detection for stability of results.

[CY]

For all I know, long time stored FFPE causes denaturation and fragmentation and shears DNA molecule but not neutralize them (make certain fragment disappear), right? So most of the DNA molecule remain diplod. I did some research and now the impression is that difficulty of de-crossing on FFPE sample makes DNA extraction and PCR uncomplete and mix up the diplod background. I am not sure if I understand it right...

[Eric T.]

The molecules don't exactly disappear, but after shearing a larger fraction of the remaining fragments will be too small (e.g. <200bp) and will be lost during wet-lab processing and sequencing, I think.

[CY]

Yep. I remembered. Guess you are right. That is one of the reason for getting noisy background.

[linouhao]

so is there any qc criteria for cnvkit of detecting wes, wgs, tr sequence, for example depth? because we often talk about low quality, but there often saying no concrete value for these. for example, you said larger bins for low quality FFPE,but what the value ,wes target bin to be 2000 instead of 267?
thanks a lot