Analysis scripts post Remora classification

[lkwhite]

I have several one-off steps and script tweaks from mid-October onwards that have yet to be incorporated into this pipeline. All of these are run after Remora classification of charged and uncharged reads:

1. Convert per-read charging information into counts per million for charged and uncharged reads. I wrote a job submission script & the generically named but well commented python script it calls (process_tRNA.py) on Bodhi at /beevol/home/whitel/data/AAtRNAseq/biological_tRNA/outputs/remora_classification/tsvs/processing to handle this.
2. Calculate basecalling error frequencies from mod bams after Remora classification and ML tag integration. This involved some minor tweaks to our existing get_bcerror_freqs.py to filter on ML>199 and then create separate entries for charged and uncharged tRNA isodecoders; the script I am now using is located at /beevol/home/whitel/data/AAtRNAseq/AAtRNAseq/src/get_bcerror_freqs_post_remora.py
3. [optional] Get per-read error information. We don't leverage this in version 1 of the preprint so tbd on whether this is necessary but we will want it built in eventually. The strategy here will work but requires tweaking to run on the mod bams after ML tag integration as I did for bcerror in 2 above.

Additionally, we should discuss whether the script filter_reads.py can be deprecated in the workflow; @MSajek I think I remember you saying this now happens as a prerequisite of inference, but I'm not sure how that squares with what we discussed with Marcus & team in December.