include process to estimate error

[ggabernet]

EstimateError can be used to estimate sequencing and PCR error rates: https://presto.readthedocs.io/en/stable/examples/tasks.html#estimating-sequencing-and-pcr-error-rates-with-umi-data

A process could be included to estimate these errors per sample

[StephenWist]

Hello!

I decided to take a shot at this issue. I've never worked on a pipeline of this complexity so I'm still figuring out how to correctly define the inputs and outputs, how everything connects, etc...

What I've managed to do so far is make a process for EstimateError.py and got the pipeline to run on a sample of the Stern2014 data. Since it's such a small subset of the data (10K reads) and I'm testing on my wimpy home computer, I did not include additional EstimateError arguments like in the example.

I can keep working on this-I think only R1 is used at the moment and I still need to get the outputs and logs going where the rest of the Presto outputs/logs go. Currently the output files are being written to work dirs.

Can you advise if I stuck the process in a sensible place? It can run on the fastq outputs of PairSeq, PostConsensus_PairSeq, ClusterSets, and Parse_ClusterSets. My idea was running EstimateError right after PairSeq is the best choice. Maybe that's not the case.

Here's the code so far if you'd like to see: Comparing changes to master branch.

[StephenWist]

Logs and outputs are now being sent to a dir like the rest of the Presto outputs, and EstimateError is run on both R1 and R2. However, R2 almost never has a UMI if I remember correctly. Maybe that will cause unnecessary running of EstimateError and a flag or check for which read has the UMI could be added.

[StephenWist]

EstimateError now runs on R1 or R2 based on the barcode_position variable: https://github.com/nf-core/airrflow/commit/c504ac6c86e77bef7d4eaa23e24cc6a2b2b007b9

Still to do (I'm putting this here mainly as notes for myself):
Double check that running EstimateError where I put it (just after Presto PairSeq UMI) makes sense from a biological perspective.
Also, check how running the pipeline with library_generation_method set to any value besides specific_pcr_umi alters the workflow in regards to EstimateError.