How Varscan process the pileup data?

[david.tamborero]

No idea, I only have used the somatic command. You can try the other Varscan commands (as the pileup2snp) with a dummy pileup file and play with the --min-avg-qual parameter to check it out.

By the way, I need that Varscan performs the quality filtering because the samtools mpileup command also does it wrong (the -Q argument does not work). Seems that the base quality filtering is not that important!

cheers,
David

[Jane M]

Hello David,
I intended to try Varscan with normal and tumoral files, so thanks for mentionning the problem. Please, keep us informed if you got news from the authors. I would like to try the tool when the problem is solved!
Do you know if the problem is also present when using Varsan with only one file?

[david.tamborero]

Just for those that can be interested in the varscan filtering according to the base quality: apparently Varscan has a bug that arises when certain characters appear in the field containing the bases (for instance, the '*'). This provokes a shift in the 'base-base quality' correspondance and then the counting of bases supporting each allele becomes wrong.

Therefore, the base quality filtering performed by Varscan fails for those reads containing these characters. I've written to the authors, hope it helps.

cheers,
david

[david.tamborero]

Thank you for your response, dkoboldt.

I indeed thought in the possibility that some 'default quality filtering' was being applied, but (for instance) in the first example, all G's should be not taken into account since their qualities are either '#' or '!'. However, Varscan stats say that for that position there are two G's. This does not seem to fit.

There is plenty of things like that, and I am not able to find a 'pattern' of quality filtering by checking the pileup positions as compared to the corresponding VArscan entry...

thanks a lot,
cheers
David

[dkoboldt]

VarScan includes a minimum base quality threshold requirement in order to count reads. By default, this is 15 or 20 depending on the command you use.

In the first example, some bases have a base quality represented by '#', indicating a Phred score of 3. In the second example, one reference base has a quality represented by ".", which is 14. Both of these fall below the minimum base quality threshold in VarScan.

[david.tamborero]

Hi,

if any of you has used Varscan to find somatic mutations given normal and tumor pileup files, maybe you can give me some light about how it processes the data.

For instance, given the following position in the tumor pileup file:

chr1	173824	T	21	GG...*..*.G.......GG.	
				##9!!!!;!<#;<<=<<<!#<

Varscan says that 8 tumor reads supports the 'T' and 2 tumor reads supports the 'G'. I do not see how it is fitted!

Another random example:

chr1	30028	c	9	,tt,,,T,,	
				5::778?.4

Varscan says that in the previous position, there are 3 reads for nucleotide 'T' and 5 for 'C' (why not 6??). There are these kind of discordances in almost all the positions I've checked, so I'm pretty confused.

thanks!
david