FilterConsensusReads: Multiple non-secondary, non-supplemental R1s

When using FilterConsensusReads with v1.5.0 I receive the following error:

Multiple non-secondary, non-supplemental R1s for Solexa-272222:1089771

Starting from a BAM file aligned with hisat2 I do the following:

1. CorrectUmis
2. GroupReadsByUmi
3. CallDuplexConsensusReads
4. GATK SamToFastq
5. align with hisat2
6. GATK MergeBamAlignment
7. FilterConsensusReads

Notably, I do not see this issue when aligning with bwa at step 5, regardless of which aligner was used to generate the initial aligned BAM. I’m not sure if this is an appropriate issue to discuss with you all, but do you have any thoughts on what may be going on here?

[roysomak4]

@nh13, I encountered the same error with FilterConsensusReads v 1.5.0 and can confirm that it was resolved by eliminating the -M flag from the bwa mem command.

[nh13]

@stkgo also take a look at this issue: https://github.com/broadinstitute/picard/issues/589

I wonder if the multiple primary alignments being output by MergeBamAlignment is also caused by the bwa mem -M option which is causing supplementary alignment to be output as secondary.

@tfenne is writing a new tool to replace MergeBamAlignment you can test here: https://github.com/fulcrumgenomics/fgbio/pull/778

Feel free to reopen if you have more questions.