From FASTQ to clean BAM using GATK tutorial #6484

Question

[lamteva.vera]

Hi!

I'm trying to build an efficient pipeline for processing amplicon sequencing data. The problem is that ValidateSamFile reveals a bunch of errors in BAM files after running BamClipper (whereas BAMs were free of errors before). Exemplary output of ValidateSamFile (MODE=SUMMARY):

HISTOGRAM   java.lang.String
Error Type  Count
ERROR:INVALID_FLAG_SUPPLEMENTARY_ALIGNMENT  138
ERROR:INVALID_MAPPING_QUALITY   315
ERROR:MISMATCH_FLAG_MATE_UNMAPPED   217
ERROR:MISMATCH_MATE_ALIGNMENT_START 8775
ERROR:MISMATCH_MATE_CIGAR_STRING    2385125
WARNING:MISSING_TAG_NM  2387464

I've read that MergeBamAlignment is a powerful tool for cleaning BAM files while preserving original read information and base quality scores. So I decided to implement the GATK's tutorial #6484 into my analysis pipeline to get rid of the errors.

I just want to ask the community's opinion about the following workflow:

I could have missed something. Any critical thoughts are welcome.

[GenoMax]

If I am reading the flow diagram right, why are you adding unaligned BAM data back into final BAM? Isn't that duplicating many reads (aligned and original copy).

[lamteva.vera]

GATK claims that

Broadly, the tool [MergeBamAlignment] merges defined information from the unmapped BAM (uBAM, step 1) with that of the aligned BAM (step 3) to conserve read data, e.g. original read information and base quality scores.

[GenoMax]

I see. Have you compared the merged BAM with the aligned BAM to see what MergeBamAlignment did?