Decouple Ensembling

[ntalluri]

I got distracted and decoupled the ensembling code #163

[ntalluri]

(Something to think about) Should I consider modularizing each part of the ML code? I can't recall if I tested each ML function individually with edge cases. There's a possibility that while HAC might work even though PCA fails when dealing with those edge cases. Right now the code will not run HAC after PCA fails due to the assumption that HAC will also fail.

[agitter]

The overall design is good. I only have minor suggestions.

Should I consider modularizing each part of the ML code?

This is low priority for me given how many other things we have to work on. If a user ever requests that or has a problem with it, we could implement it.

[agitter]

I don't have any major comments remaining. I'll test locally and merge if everything looks good.

[agitter]

The tests passed locally. However, when I ran with the config file, I was confused about the the data1 outputs. Should we have ML outputs (e.g. PCA output) for Omics Integrator 1? We run it with 4 parameter combinations and the summary table shows:

output/data1-omicsintegrator1-params-E3LSEZQ/pathway.txt	3	2	1	2	2	1	1
output/data1-omicsintegrator1-params-NFIPHUX/pathway.txt	0	0	0	0	0	0	0
output/data1-omicsintegrator1-params-SU2S63Y/pathway.txt	3	2	1	2	2	1	1
output/data1-omicsintegrator1-params-V26JBGX/pathway.txt	0	0	0	0	0	0	0

I didn't get those outputs.

[ntalluri]

When I run OI1 locally, I am getting back outputs for the ml step. I keep rerunning oi1 for the config file and seem to be getting back what I expect.

[agitter]

When I ran it, I thought I did not get those OI1 ML outputs. However, I tested it again and I did. I could have been mistaken or had some temporary error.