Enquiry on consistent segments amplified/deleted across most samples

[michelle-jahja]

Hi, I was wondering if this is an issue that has been seen before. I am new to facets and copy number calling so I would appreciate any help/advice.

I have been running FACETS for several samples (WES and targeted sequencing with matched normals) from a public dataset. I noticed that in these samples, certain chromomes/ certain regions in chromosomes were weirdly amplified or deleted - looking at the log-ratio, it does not seem clear that these segments should be 'amplified'. I have attached images as examples.

These regions (chromosomes 16, 17, 19) are called as amplifications throughout almost all samples (~170 samples), which 1) biologically does not make sense as these chromosomes are usually not amplified in these cases, 2) based on karyotype information of each sample, these chromosomes should not be amplified.

I was wondering if this is a sample processing issue, or if there is a parameter in facets that can control for/account for these slight shifts in log ratio that shouldn't be called amplifications?

Could it be due to different read depths between matched normals vs tumor samples? I have also attached histograms of tumor and normal read depth of one sample - although this difference is not seen in all samples.

Here are the current parameters for running facets on WES and targeted sequencing.
# for WES xx <- preProcSample(rcmat, gbuild = "hg38", ndepth=35, cval = 150, snp.nbhd = 250) oo <- procSample(xx, cval=500)

# for targeted xx <- preProcSample(rcmat, gbuild = "hg38", ndepth=35, cval = 150, snp.nbhd = 150) oo <- procSample(xx, cval=500)

Thank you and very grateful for any thoughts/opinions!

[veseshan]

This seems like sequencing artefact. The log-ratio for total copy number is clearly different but it is such a low level and shows no allelic imbalance. Wonder why tumor alone has higher coverage than normal in those chromosomes (16, 17 and 19).

[michelle-jahja]

Thanks for the response, considering that these are likely sequencing artefacts, do you have any suggestions on how I could account for these when visualising my data? Is it right to call these as 'diploid' or 'artefact' with the condition that the log-odds ratio/mafR is < a certain number? Happy to hear your thoughts - thanks!

[veseshan]

You can look at the variation in total coverage (normalized to say 1 million reads) in the normals by chromosome. That can provide some idea of the expected variation.