Optimize attachBarcodes in sctools

[kbergin]

Why?

AttachBarcodes is one of the longest steps in the Optimus pipeline. 3.5 hours on PBMC4k. 6 hours on Bo Li’s test dataset for the cumulus package. We need to make the step faster.

The code can probably be improved, but it is also theoretically parallelizable.

Where to start?

The related code's entry is probably here, from [~accountid:557058:32a3563d-4613-4e1b-a777-0770d66a2e0c]. Ambrose Carr may have ideas on how to improve this step. There is the possibility of error rate related bottleneck that should be investigated.

AC:

• AttachBarcodes is faster and does not change the outputs scientifically
• Update tests in sctools, add if needed
• Release sctools

It’s possible the below can go to a new ticket

• Update Optimus version of sctools
• Coordinate with Jodi to update documentation
• Update pipeline tests if needed

┆Issue is synchronized with this Jira Dev Task