UMI retention with single-end reads bug

[anoronh4]

Under specific conditions, the UMI sequence is removed by the STAR aligner. In this bug the reads must have the read designated with a /, e.g.:

@SRR5665260.1.36873006/1

and the sample must also be single-ended and have UMI. After extraction using umi_tools extract, the UMI appears as follows:

@SRR5665260.1.36873006/1_CAG

but STAR reads up to the / and discards the rest, such that the UMI barcode that follows is discarded. if umi_tools extract is used on paired end reads the read ID is printed differently:

@SRR5665260.1.36873006_CAG

and is therefore properly parsed by STAR.

We could add an extra process to fix the read IDs but this would take extra storage and is probably better handled with a proper bug fix in umi_tools. this is not urgent as single-ended reads are not a common use case for us and the /1 suffix is similarly rare. we can put it on the backburner for now but i opened an issue for it on the UMI-tools github