Hi all. When using the umi workflow and fgbio consensus calling, is it expected that the downstream samtools etc tasks still run (eg

FASTQ_PREPROCESS_GATK:CONVERT_FASTQ_UMI:SAMTOOLS_VIEW_MAP_MAP

) before going to realignment ? For me at least, that means all reads will be unaligned reads (consequence of fgbio consensus calling) , which eventually get merged (cat) together with empty aligned reads before the downstream BWA realignment.

Sounds like a bug. Frankly the fgbio consensus section of the pipeline is not very well tested or maintained.

Yep . Looking a bit more now. Seems to be not a lot of point running those tasks apart from getting the bam into fastq format before bwa realignment. Doesn't actually do anything as all reads (all are unaligned) get incorporated at the merge stage and pushed to realignment. I may consider use of fgbio FilterConsensusReads after the realignment step and turning most of the samtools stuff off post consensus calling. I've already got a few additions to the umi workflow in our pipeline customization/config , including additional custom trimming with cutadapt I've brought up here before. Would be good to know who is using this to get a consensus on consensus calling workflow.

I don't think it is well used. Feel free to contribute 😉

Yep, happy to try and contribute - i've suggested that here previously. May be a little tricky as our customizations are quite a few , so would need to separate out such an element (i'd need to look back to see if suitable stand-alone commits/branches in our forked repo).

To be honest, my recommendation would be to use nf-core/fastquorum

May be operationally convoluted for us, but i'll look. I think a few more tweaks to sarek should bring up to umi/fgbio best practise .

well, I think fgumi is better practice now

For me in a perfect world, #C0453Q2SFCM should be the most up to date pipeline to deal with UMI, I do believe it is slightly out of data as well unfortunately. And we would chain this with #CGFUX04HZ

I'm hoping that Nils manages to include a

fgumi runall

command (github.com/fg-labs/bwa-mem3/issues/196#…) that pipes together all the consensus building steps into one, that'd be much easier to include into e.g. sarek

oh yeah

well, may still have uses 😉