Dear all,
I do CNV analysis. I use Affymetrix Cytoscape 750K array. There are 14 probes inside a genomic region chr7:126.229.425-126.308.312 (half of them - CGH probes for CNV detection, half - SNP probes). And these probes indicate a presence of a CNV.
We also have analysis of CNVs in the same sample in shallow NGS data of 4x. The usual resolution where we can guarantee the detection of CNV in shallow WGS of this depth is 25kbps. The variant is much bigger - 70kbps.
We don't see any signal in the abovementioned region. Coverage is exactly equal to diploid - somewhat higher, somewhat lower in different parts. Mosaic variant can usually be seen by moderate increase/decrease of coverage - it is not the case.
As a comparison for aCGH analysis we've 30 non-related samples, so the presence of duplication there may be excluded.
The array sample is not very noisy. Noisy, yes, - but the drop of signal there is absolutely clear and can not be explained by the noisyness. Sample swap is not the case - the other rare variants match perfectly.
Does anyone know how it may occur? We had several such cases, idk, 4 CNVs per 300 analysed pairs array/NGS.
That's how array intensity looks like (+region around the CNV, only aCGH probes are shown) 
This sample was sequenced with shallow NGS twice. Each dot at the IGV plot is 5kb coverage. In theory, in the center there should be a drop of 14 dots - visible drop. But nothing.
My guess would be that there are variants there preventing the CGH probes from binding.
Thanks a lot, this is a really good version! However there are 7 probes inside (I just showed CGH ones, but there are 7 more that is used by Cytoscape software - can not visualize them due to lack of answer to this question C: Snp array intensity ) - and it has to be very unfortunate to have variants in all these probes...
SNPs blocking the probes is more likely than an issue with the NGS.
This one i dont understand. I agree on case of amplicon sequencing - one per 1000 amplicons will not bind, but this is true seq pcr free, it is 70 kb deletion and in NGS I see absence of coverage drop - in case of incorrect mapping i would expect drop of coverage, not stable one.
NGS is less prone to mapping errors due to variants that would affect probe binding.
Sorry, sorry, got it wrong,i am not a native speaker. So, it is likely a false positive of array? All my colleagues are saying this, but I am confused by the strength of the signal...
While it appears to be so you would want to try something independent to be 100% sure. Especially since we are in a Yes:No situation.
Will merge two shallow genomes together and run Manta with relaxed thresholds...hopefully will be enough - people won't do mlpa or qpcr for this sample...
Hopefully that clarifies things. A different method would definitely be preferred though.
No evidence of sv in Manta analysis. One inversion was found inside the region, but super small. So, I guess, this is an artifact, despite such a strong signal.