Strand Bias filtering

c_ro87 · 04-04-2014, 07:41 AM

Hello all,
i run the variantCaller in germline_low_stringency mode with the default parameters

the strand bias threshold for snp Strand Bias is 0.95, everything up that value is considered strand bias

we did some calculations using Relative Strand Bias Formula
http://mendel.iontorrent.com/ion-doc...0/31621166.jpg
We found some cases where this filter works bad, for instance when we have FSAF:2 FSAR:1 FSRR:10 FSRF:0, the STB is 0.916 and therefore accepted as variant. Or FSAF:1 FSAR:1 FSRR:20 FSRF:20 the STB is 0.5 and therefore accepted again. So... we accept that calls???

arnaud83 · 04-12-2014, 01:21 AM

Quote:

Originally Posted by c_ro87

Hello all,
i run the variantCaller in germline_low_stringency mode with the default parameters

the strand bias threshold for snp Strand Bias is 0.95, everything up that value is considered strand bias

we did some calculations using Relative Strand Bias Formula
http://mendel.iontorrent.com/ion-doc...0/31621166.jpg
We found some cases where this filter works bad, for instance when we have FSAF:2 FSAR:1 FSRR:10 FSRF:0, the STB is 0.916 and therefore accepted as variant. Or FSAF:1 FSAR:1 FSRR:20 FSRF:20 the STB is 0.5 and therefore accepted again. So... we accept that calls???

Strand bias is not the only parameter to take into account to remove false positive. Alternate allele frequency, Alternate allele count, Mapping quality, position on the read can be very usefull to filter variants.

IonTom · 04-13-2014, 04:45 AM

For Amplicon data a strand bias seems to be much more frequent than for Exome sequencing. So maybe filters should not be quite as strict. Maybe it is enough to check if a
variant is found on both strands, or at least one percent on each strand. On the other hand
i analyzed one sample where there was a significant bias for amplicon data and the strand bias was also found on exome data for the same sample.

The Mutascope publication (specifically the supplement) discusses the Amplicon strand bias problem in more detail: http://bioinformatics.oxfordjournals...rmatics.btt305

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Strand bias reads	Nino	Bioinformatics	4	12-12-2013 08:38 AM
strand bias P-value	lyz1030	Bioinformatics	1	07-16-2012 01:12 AM
strand bias	zorph	Bioinformatics	0	06-13-2012 03:44 PM
Evidence of Strand Bias	AnamikaDarwin	Bioinformatics	1	06-27-2011 05:21 PM
Forward strand bias?	dawe	Bioinformatics	6	04-05-2010 08:00 AM

04-04-2014, 07:41 AM	#1
c_ro87 Member Location: Buenos Aires Join Date: Feb 2012 Posts: 12	Strand Bias filtering Hello all, i run the variantCaller in germline_low_stringency mode with the default parameters the strand bias threshold for snp Strand Bias is 0.95, everything up that value is considered strand bias we did some calculations using Relative Strand Bias Formula http://mendel.iontorrent.com/ion-doc...0/31621166.jpg We found some cases where this filter works bad, for instance when we have FSAF:2 FSAR:1 FSRR:10 FSRF:0, the STB is 0.916 and therefore accepted as variant. Or FSAF:1 FSAR:1 FSRR:20 FSRF:20 the STB is 0.5 and therefore accepted again. So... we accept that calls???

04-13-2014, 04:45 AM	#3
IonTom Member Location: Germany Join Date: Apr 2014 Posts: 32	For Amplicon data a strand bias seems to be much more frequent than for Exome sequencing. So maybe filters should not be quite as strict. Maybe it is enough to check if a variant is found on both strands, or at least one percent on each strand. On the other hand i analyzed one sample where there was a significant bias for amplicon data and the strand bias was also found on exome data for the same sample. The Mutascope publication (specifically the supplement) discusses the Amplicon strand bias problem in more detail: http://bioinformatics.oxfordjournals...rmatics.btt305