Who is the culprit? Picard, Samtools or GATK?

[newbietonextgen]

hi,

I am using RNA-Seq data and my work flow is below..and am having trouble with BAM file

Single end fastq->Shrimp alignment->sam file->changed to BAM using Samtools->added read groups using Picard->index and Now fed this into GATK-> works fine.

In best practice of GATK v3, they ask you to combine all your BAM files. So i took the bam files after read group were added using Picard -> Picard Mark duplicates->Mergesamfiles(BAM can be input too)->index the file and the Validate BAM file-> result (No errors found). I did not get any error in all these steps.

Now the combined BAM files fail to be recognized by GATK throwing an error "ERROR MESSAGE: Bad input: Could not find any usable data in the input BAM file(s)."

I looked at the BAM files using IGV viewer and it looks fine and all the read have read groups and quality of 255.

1) so why is GATK doing it? Is there a problem with my BAM? Any ideas?


I was reading in GATK forum and found an explaination as to why it happens by a user " In my experience, some mapping program do not calculate map Quality score correctly, it will output 255 as mapping score, but 255 actually in GATK's CountCovariatesWalker is filtered out by MappingQualityUnavailableFilter in the top of walker:
@ReadFilters({MappingQualityZeroFilter.class, MappingQualityUnavailableFilter.class})
just get rid of MappingQualityUnavailableFilter.class in this line, then it will works...(this works for my case..)"

2) HAs anybody noticed it?

Thanks

[mateidavid]

Did SHRiMP compute mapping qualities? Look in an output file, if you see any MQV of 255, then the answer is NO.

If the answer is YES, some program down the line must have lost them somehow.

If the answer is NO, you have to rerun the mappings. Find out which option disabled MQVs in SHRiMP, and disable it or work around it.

[newbietonextgen]

Shrimp worked well. I thinks its some program down the line that goofed the results. For now i just am using a different approach (Freebayes) and no problems yet..