BWA and Solid data issue

[ulz_peter]

Hi all,

We recently got Solid whole exome data and I tried to align it with bwa and do the snp calling with GATK.
After resolving the CS/CQ-tag issue by a python script to adapt the bam file for analysis with GATK I now ran into another problem:
Alignment worked well and SNP calling did not show any error messages but: It seems that there is some trouble with the phred quality of the bases. GATK showed me some SNPs with AF=1.00 (so homozygous) which were obviously heterozygous. IGV shows the Phred score for each position and it seems that every base which mathces the reference sequence has phred score 0 whereas the SNPs have a relaitvely high Phred Score (>25).

Is it possible that bwa changes the base quality scores when converting the csfasta and qual files to the bwa double encoded fastq files? Or is there a special option in the GATK for Solid base-call qualities? Am I doing something completely wrong?

Any help is appreciateed...

[nilshomer]

It uses the MAQ formula, which is also used by BFAST. See the section called "ABI SOLiD Base Qualities" in the following page for an explanation: http://sourceforge.net/apps/mediawik...apping_Quality. You can confirm by checking the function "cs2nt_nt_qual" in "cs2nt.c" in the BWA source code.

[ulz_peter]

Hi Nolshomer,

Thank you for the answer. I checked the resulting SAM-File and it appears that the quality values are distributed nicely within 0 and 60 (substracting 33) but with most of them being at 60. So it seems as bwa does a good job...
Does anyone know of a switch for the GATK for Solid data? And why does IGV show so many zero quality bases where there shouldn't be too many?

[ulz_peter]

Sorry for the misspelling of your name Nils :-)

[ulz_peter]

Just a short update:
when looking at the details in IGV the OQ (original quality)-Tag shows more or less only '!' (ASCII:33) characters. The CQ-Tag seems to be ok. After converting the duplicate marked, realigned and recalibrated bam file back to SAM Format, it seems that most of the qualitites were changed to '!'. I checked the intermediary bam files (after marking duplicates and realignment) and they showed nice quality values again. So I guess that's an issue of the GATK recalibrator. The weird thing is that the recalibrator obviously does not write the original qualities in the OQ tag...

Has anyone seen this before?

[Brugger]

We are using the GATK for post-mapping analysis, but do not recalibrate.

One problem we have seen if that the solid reads will leak back to the reference if there is one colour error, and this will screw the splits of bases at a snp. It can be so bad that a homozygous snp becomes +30% reference, I have a solution that scrubs away the potentially wrongly corrected bases at SNPs and we are getting nice results with this approach.

[zlu]

We are using the GATK for post-mapping analysis, but do not recalibrate.

One problem we have seen if that the solid reads will leak back to the reference if there is one colour error, and this will screw the splits of bases at a snp. It can be so bad that a homozygous snp becomes +30% reference, I have a solution that scrubs away the potentially wrongly corrected bases at SNPs and we are getting nice results with this approach.

Hi Brugger, do you mind sharing your solution? Does this still involve recalibration using GATK? Thanks.

[aldo]

Hi Brugger,

I am also very interested in the approach you have taken. Could you share it with us? Thanks

[Brugger]

My approach looks at the original colours and make some corrections based on the observations. For this to work you need to use a mapper that retains the original colour information like bioscope. Furthermore, I have a patch for bwa that does just this as well.

I can see if I can dig out some stats/graphs showing the post mprovement.

I have not looked at this for a while, but it was the plan to make it public available when I have some spare time.

If you want to test this please send me a PM with your email address and we can sort something out.