How to create duplex UMIs with AnnotateBamWithUmis

I am working with data that uses two UMIs for paired end reads. One UMI was included as part of index 1 and the other as part of index 2. I’d like to annotate the RX field in my BAM file with both UMIs with a dash between, as in NNNNNNNN-NNNNNNNN. I see that CorrectUmis can handle duplex UMIs, such that it looks for the consensus sequence independently for each half. What would be the best way annotate a BAM with duplex UMIs in this case? It seems with AnnotateBamWithUmis you end up replacing the contents of the RX field when you run it twice.

[birnbera]

Got it. My use case is different but it makes sense to do more aggressive quality filtering for variant detection. Thank you for your help on this!

[nh13]

UmiAwareMarkDuplicatesWithMateCigar is mainly used to ensure that two or more reads from the same source molecule are not counted twice in downstream variant calling, but why not leverage the multiple observations for variant calling. I am mostly interested in using UMIs to detect groups where reads that original from the same source molecule are in the same group so that I can then create a consensus from each group of reads. The consensus has much lower error rate and sources of bias (depending on the UMI scheme) so can enable ultra-accurate variant calling (SNVs/indels/STRs/etc.).

In the case of GroupReadsWithUmi it so aggressively filters out read.

Well you can turn off filters and some additional hard-coded filters are being relaxed (see #648) but we’d welcome an opportunity to collaborate.