Filtering steps for VarScan2 somatic and germline calls

[ElenaN]

What is the proper filtering pipeline for somatic mutation detection using VarScan2, with tumor/normal paired samples?
My understanding of the steps:
0. VarScan2 somatic to make calls
Then filter:
1. VarScan2 processSomatic to separte calls Germline/Somatic/LOH and call some "high confidence"
2. VarScan2 somaticFilter on the Somatic.hc file to do further filtering.
- as I understand, this is where filtering for false positives aroudn indels take place
- it also seems to look at p-values and such - is it not redundant with the previous step?
3. fpfilter.pl on the resulting data for further filtering?
(in my hands this step seems to hang indefinitely without doing anything, though. not sure why)

My questions are:
- is this a correct order of steps, or are some of the 3 filtering steps redundant or obsolete?
- what are the recommended parameter settings for bam-read-count, especially mapping quality q? Should it be 1, as is recommended for mpileup, or higher?

Now, if I want to detect germline mutations in the same sample, can I just use the "Germline" (and I guess LOH) calls from VarScan?
Dan Koboldt seems to think so, as long as one uses the normal bamcount in step 3. Should "VarScan2 filter" be used in step 2? Or should a separate germline mutation caller be used for this purpose?

Thank you,
Elena

[dkoboldt]

Elena,

You could probably just perform steps 1 and 3 and get a good high-confidence set of somatic mutation calls.... the somaticFilter command is more simplistic than the filter false positives.

If you want the germline calls, combine Germline.hc and LOH.hc files, and then run that through the false positive filter with the NORMAL bam.

As for bam-readcount, I'd recommend at least mapping quality of 1 and base quality of 15.

[ElenaN]

Thanks, Dan!
(BTW, I just posted another VarScan question as a separate thread http://seqanswers.com/forums/showthr...364#post109364 - it'd be really great if you could answer it as well.)

[zhixiang]

Elena,

You could probably just perform steps 1 and 3 and get a good high-confidence set of somatic mutation calls.... the somaticFilter command is more simplistic than the filter false positives.

If you want the germline calls, combine Germline.hc and LOH.hc files, and then run that through the false positive filter with the NORMAL bam.

As for bam-readcount, I'd recommend at least mapping quality of 1 and base quality of 15.

Hi dkoboldt,

Can filterfp.pl be used to filter the somatic indels?

Can I combine the indel vcf and snp vcf from the output of varscan somatic, and then filter the vcf file using step 1 and step 3? Does this works for both snp and indel in the vcf file?

[thek71]

Hi all,
I have a similar question regarding the germline mutations. I have several tumor-normal samples for which I have used the somatic command and I want to check several genes for germline mutations. From the publication is stated that hc variants are the ones with at least 10% allele frequency in normal and tumor samples.
This varinat is in the hc file:
chr1 565088 C T 0 37 100% T 4 19 82.61% T Germline 1.0 0.018159073897484498 2 2 10 9 0 0 22 15
while this one
chr1 741267 T C 33 38 53.52% Y 30 22 42.31% Y Germline 5.540944545408712E-23 0.9211526256316275 2 28 3 19 10 23 5
33
is not. The numbers of reads and the percentages seem to be ok, but I fail to understand why the variants are categorized differently.
Any help would be highly appreciated.

Thank you in advance