Bfast color space tag and GATK error

[adamhfreedman]

The Genome Analysis Toolkit 'CountCovariates' portion of the base quality recalibration tool utilizes platform information, and in the case of color space data, reads the colorspace (CS) tag when working with SOLiD data. From an alignment generated with Bfast, I am getting an error using GATK:

'Unrecognized color space in SOLID read, color = . Unfortunately this bam file can not be recalibrated without full color space information because of potential reference bias.'

As far as I can tell, this occurs because for GATK, only 0,1,2 and 3 are allowable states, but Bfast also includes, '.'

An example SOLiD tag from Bfast:
'CS:Z:T033.0330033003300123.22122221212102233223313322131'

And...from the GATK source:

' private static byte getNextBaseFromColor( SAMRecord read, final byte prevBase, final byte color ) {
switch(color) {
case '0':
return prevBase;
case '1':
return performColorOne( prevBase );
case '2':
return performColorTwo( prevBase );
case '3':
return performColorThree( prevBase );
default:
throw new UserException.MalformedBam(read, "Unrecognized color space in SOLID read, color = " + (char)color +
" Unfortunately this bam file can not be recalibrated without full color space information because of potential reference bias.");'

Does anyone have a workaround for this problem?

Thanks,
Adam

[nilshomer]

Originally Posted by adamhfreedman

The Genome Analysis Toolkit 'CountCovariates' portion of the base quality recalibration tool utilizes platform information, and in the case of color space data, reads the colorspace (CS) tag when working with SOLiD data. From an alignment generated with Bfast, I am getting an error using GATK:

'Unrecognized color space in SOLID read, color = . Unfortunately this bam file can not be recalibrated without full color space information because of potential reference bias.'

As far as I can tell, this occurs because for GATK, only 0,1,2 and 3 are allowable states, but Bfast also includes, '.'

An example SOLiD tag from Bfast:
'CS:Z:T033.0330033003300123.22122221212102233223313322131'

And...from the GATK source:

' private static byte getNextBaseFromColor( SAMRecord read, final byte prevBase, final byte color ) {
switch(color) {
case '0':
return prevBase;
case '1':
return performColorOne( prevBase );
case '2':
return performColorTwo( prevBase );
case '3':
return performColorThree( prevBase );
default:
throw new UserException.MalformedBam(read, "Unrecognized color space in SOLID read, color = " + (char)color +
" Unfortunately this bam file can not be recalibrated without full color space information because of potential reference bias.");'

Does anyone have a workaround for this problem?

Thanks,
Adam

You could replace all "."s in the CS tag with a random color. Not the correct solution, but it should work.

[sdvie]

Dear all, sorry for reactivating this thread after a while, but I thought this fits in here:

In my experiment, some colorspace sequences turned out to contain "4" after aligning them with bfast:

in the fastq file (generated by the bfast solid2fastq script), one example looks like this (obviously bad quality read, but please read on):

@1386_370_11
T21......3....02...302012002031112020320.3212..01..
+
16!!!!!!8!!!!1?!!!;=>&'&3.2%(:>3%'6>-&%!1&&)!!,&!!

the reads were aligned using
- bfast bwaaln
- bfast localalign
- bfast postprocess

and the same read in the .sam output (excerpt, columns 1, 17 and 18) looks as follows:

1386_370_11	CS:Z:T21444444344440244430201200203111202032043212440144	
CQ:Z:16!!!!!!8!!!!1?!!!;=>&'&3.2%(:>3%'6>-&%!1&&)!!,&!!

not quite clear how this happens.

In any case it causes trouble when trying to run the Covariates Recalibration tool from GATK which legitimately complains about the "non-base characters" in the sequence:

java -jar /soft/bio/sequence/GenomeAnalysisTK-1.0.5506/GenomeAnalysisTK.jar -T CountCovariates
##### ERROR MESSAGE: SAM/BAM file SAMFileReader{/home/sderdak/temp/BM3062_nodup_realigner_csadjusted_readgroups.bam} 
is malformed: Unrecognized color space in SOLID read, color = 4 Unfortunately this bam file can not be recalibrated 
without full color space information because of potential reference bias.

Any explanation for the "4's" will be very much appreciated, thank you.

[nilshomer]

All dots in the input will be converted to 4s in the output. 4s are ambiguous colors, like Ns are ambiguous bases.

[sdvie]

Originally Posted by nilshomer

All dots in the input will be converted to 4s in the output. 4s are ambiguous colors, like Ns are ambiguous bases.

ok, good to know.
Thank you Nils.

[nexgengirl]

GATK now has a patch for this:

http://gatkforums.broadinstitute.org...d-color#latest