caporaso-lab / pretrained-feature-classifiers Goto Github PK

These are leftover from the pre-rescript days, and aren't being used anymore in this project, to the best of my knowledge.

Edit, I forgot that #10 was never completed - that's what I had in mind when I opened this issue, the original intent of that PR was to clean up the data fetching steps, among other things but I see now that a) that work was never finished and b) the work that was completed was never merged.

cc @colinbrislawn

I just realized that the most resent SILVA files provided on the 2021.4 Data Resources page are still pulling and compiling from the SILVA 138 and not the 138.1 version of SILVA (according to the provenance information). We just need to update the pointer to the 138.1 version which is already handled through get-silva-data.

Just for a quick reference, I think we only need to update these lines (thanks @nbokulich!) :

• https://github.com/caporaso-lab/pretrained-feature-classifiers/blob/master/Makefile#L24
• maybe also this line to just adopt the version number in the filename: https://github.com/caporaso-lab/pretrained-feature-classifiers/blob/master/Makefile#L27
• https://github.com/caporaso-lab/pretrained-feature-classifiers/blob/master/Makefile#L12

This should be as simple as reordering the makefile directive. This is for the same reasons as outlined in #3.

Full-length steps should be submitted before the 515-806 steps, in case the submission order needs to be used as a tie-breaker (we want the longest jobs running soonest).

• getgg and getsilva should download raw data into the inputs dir
• getgg and getsilva dump their imported data into the wrong directory

We are requesting too much time and memory for several steps (importing, verifying, etc) --- we should drop in a few more config vars to the script and Makefile to fine tune more.