high read depth and threshold for filtering

Hello!

I am working with dna-seq data from tumor samples and I have tried to plot the frequency of my read depth. I am wondering how to choose a good threshold for the read depth in order to remove variant with low depth. Furthermore I have seen that there a some variants with a very high read depth (over 600) how is it possible? what does it mean?

Thank you in advance,

Best regards

read depth

• 149 views • link

updated 9.5 years ago by Brian Bushnell

• written 9.5 years ago by noeD

1 answer

From mapped reads, you can calculate the read depth frequency using the BBMap package like this:

pileup.sh in=mapped.bam covhist=covhist.txt

...then plot the two columns using Excel or R or whatever. But, that will not give you a good idea of how to set cutoffs. Variants and coverage could occur at any depth in tumors. If you want to find causal variants that are present in virtually all the cells in the tumor, it's standard to use matched tumor/normal samples; you can't simply find them by looking at tumor data with a depth cutoff. This is a complex problem and I suggest you study the literature in much more detail before beginning analysis.

• 0 views • link

written 9.5 years ago by Brian Bushnell

Thank you for your reply and help!

I have tumor/normal samples and I tried several tool like Mutect2 (SNVs and INDEL), Varscan2 (as mutect2), Mutect1 (SNVs) for the variant calling and I have seen that them output (variants) are very different from each other, they have in common only a few variant. Could you suggest a good tool for variant calling? I have read several paper about variant caller but I didn't find the "best tool" or the best way to combine them output.

However I asked about read depth only for removing variant that are not supported by a high number of reads: for example I have variants that are supported by 10 reads, I guess that it is not enough... I have only found that the authors of this paper http://bmcbioinformatics.biomedcentral.com/articles/10.1186/1471-2105-15-125 suggest to filter the variants with a DP<=8. Is it a good threshold?

You wrote: "Variants and coverage could occur at any depth in tumors" so it is possible/normal to have a variant that is supported by 600 read? If yes, why? I apologise for my basic question, but I am new to sequencing world :)

thank you in advance!

• 0 views • link

written 9.5 years ago by noeD

You wrote: "Variants and coverage could occur at any depth in tumors" so it is possible/normal to have a variant that is supported by 600 read? If yes, why?

Yes, why not? First off, perhaps you sequenced at 600x - in that case, all homozygous variants would be seen at 600x even in normal tissue. Since you did not state your sequencing depth, there's no way to determine whether 600x is normal or not. Second, even if you sequenced at lower depth - say, 20x - some features (such as mitochondria) will still be present at much higher depth. Third, even for features expected to be present at 20x, tumors are no longer the same as the host organism. The host might have 2 copies of a gene, while the tumor might have 100 copies, particularly if that is linked to what's causing the tumor to out-compete healthy tissue.

• 0 views • link

written 9.5 years ago by Brian Bushnell