Protein Function Inference via Artificial Intelligence On a molecular scale, most diseases are caused by the malfunction of proteins. This makes understanding every proteins function crucial for understanding disease. Unfortunately, measureing the funciton of a protein is extremley difficult. To inferr function, protien-protein interactions (PPI) are used along with a few methods of machine learning.

Setting up the env.
Recommendations:
Linux
GPU
Languages:
Python
Perl
YAML
Bash
Dependencys:
Tensorflow
Numpy
SKLearn
Matplotlib

Obtaining The PPI All data files missing because they are to large to store on github. Luckily they are easily obtainable. Follow steps below to create an up to data version of them.

1. Got to https://string-db.org/cgi/download.pl?sessionId=40Cx3XVALsca&species_text=9606
2. Download 9606.protein.links.full.v10.5.txt.gz and uncompress. (this is human ppi data)
3. mkdir 'InteractionData', then move the ppi to this folder.
4. cat InteractionData/9606.protein.links.full.v10.5.txt | perl -wan PPI_parse.pl > Protein-Protein_Combined-Interactions.txt
5. Download and decompress human protein annotations from here: "GO Protein Annotations for Human"
6. Parse only the experimentally determined annotations out of tsv file using: cat "file" | perl -wane 'comming soon' > 9606_experimental.tsv
7. mkdir data
8. Using Jupyter notebook, run preprocess_data_v3.ipynb.
9. The PPI and GO will appear as serialzed numpy binary arrays in the data folder.