Re-write metagenome_contributions.py

[gavinmdouglas]

metagenome_contributions.py no longer in this repository - there could be better way to output the contributions that could leverage the likelihoods outputted in R by the discrete hidden-state prediction methods.

[gavinmdouglas]

metagenome_pipeline.py outputs functions stratified by functional contributions by default now. per_sample_functions.py is still experimental script to use probability distributions rather than discrete predictions.

[jjmmii]

Hi @gavinmdouglas , I updated the local clone today and noticed that metagenome_pipeline.py is taking a long time to run (job has been running for 8 hours and still running). I guess it's because it's calculating stratified output? Is there a way to turn off this option? Thank you so much.

Best,
-Jamie

[gavinmdouglas]

Hey Jamie,

This is definitely a problem, thanks for pointing this out. I have re-written how the stratified data is output and it is much faster now. I haven't added an option yet for non-stratified output only.

Thanks,

Gavin

[jjmmii]

Thanks so much Gavin! It is blazing fast now, but I noticed there is much less number of lines in the pred_metagenome_unstrat.tsv compared to the OUT_PREFIX.genefamilies.biom.tsv in a previous version running with the same data (853 lines vs 3333 lines respectively). Also strangely, in pred_metagenome_strat.tsv, when I check which sequences are mapped to the EC's, only a few (9 out of 485 sequences) are used/output. For example:

$ cut -f2 pred_metagenome_strat.tsv | sort-uniq-count-rank
595     seq_16
549     seq_13
504     seq_11
455     seq_9
454     seq_4
454     seq_6
443     seq_2
442     seq_5
160     seq_7
1       sequence

Coincidentally these sequences are the very first ones in my data. Could this be a bug (i.e. not all output was written) or PICRUSt2 only mapped a few of my sequences to genes?

Best,
Jamie

[jjmmii]

Just to follow-up, the problem was gone as of the latest clone of PICRUSt2 yesterday.