feat: compartment curation about yeast-gem

[edkerk]

RAVEN has functionality to run and parse WoLF PSORT. I've never tried it, and it runs on unix only, but could be useful. Additional predictLocalization function can actually assign localization, trying to reduce the number of transport reactions that need to be added.

[BenjaSanchez]

@CarlMalina which database were you using for finding proper localization of enzymes in mitochondria? We could use the same one but for all compartments in the cell

[CarlMalina]

@BenjaSanchez I have mostly been using MitoMiner (http://mitominer.mrc-mbu.cam.ac.uk/release-4.0/begin.do) which combines experimental evidence for protein localization with GO and UniProt annotation. However, it is quite specific for mitochondrial proteins so I am not sure it is of much use for other compartments.

[ChristianLieven]

I've used PSORTb (http://www.psort.org/) for my reconstruction, and I was pretty happy with the output. However, it is exclusive for prokaryotes but there is https://wolfpsort.hgc.jp/ for eukaryotes.

Not sure how that performs though.

[simas232]

RAVEN also supports CELLO. It is online-based tool, so you run it online, save the text output in your computer and then use it as input to assign localization. The calculation scores can then be used in predictLocalization. The advantage of CELLO is that it considers multiple criteria when calculating compartment scores (Nearest Neighbour, SVM) rather than relying only on Nearest Neighbour algorithm like PSORT.

[edkerk]

When considering this, take into account that mitochondrial enzymes have been manually curated in mitoYeast-GEM and will be applied as part of #12.