[Assessment] Ensure no strand-bias issues from using UMI-collapsed reads in TNscope

[mathiasbio]

Description

As we are planning to start using the UMIs in the standard TGA workflow which creates singletons out of read-pairs when the mates are overlapping, it is important to ensure that strand-bias filtering in TNscope is not an issue.

I think that at the moment if there are multiple read-pairs for a UMI group then it will decide the final strandedness of the UMI-collapsed read to be the most common strand of the original reads. This should theoretically mean that we would on average have the same strand-representation in the final collapsed bam as before which would be nice.

However, the reads are singletons, and it could be that TNscope doesn't handle them properly with regards to strandedness. I believe I had seen something like that before when doing work on this issue: #1336

So we just need to understand properly how strandedness is handled in TNscope downstream of having UMI-collapsed bamfiles before we can be safe in using it in the TGA workflow.

Suggested approach

No response

Criteria

No response

Origin

No response

Anything else?

No response