import sys
import os
import networkx as nx
import pandas as pd

sys.path.extend(os.getcwd())
from gp_tools.methods.randomwalk import RandomWalk
from gp_tools.methods.node2vec import Node2Vec

# import a protein-protein interaction graph
pp = nx.read_edgelist('./project_data/graphs/ppiall.edgelist')

# Seed genes for 
seed_genes = pd.read_csv('./project_data/data/seed_genes.tsv', sep='\t')
seed_nodes = list(seed_genes.seeds)

#############################
# Application of RandomWalk 


algorithm = RandomWalk()
print(algorithm.get_params()) # To see a dict of param you can set for this algorithm
# params = algorithm.get_params()
# algorithm.set_params(**params) # You can customize params here, you can also do this when you initiate the algorithm

algorithm.run(G=pp, seed_nodes=seed_nodes)

# RandomWalk algorithm give each node a score
print(algorithm.results) # dict of {gene id: score}
print(algorithm.get_results_df()) # datafrom with gene id and score

####################################
# Application of Node2Vec is similar
# But the node2vec is a multi-step algorithm (Please consult the node2vec citation for more details)
# Step 1: Create walks
# Step 2: Learn embeddings based on the walks
# Step 3: Calculate the similarity score to seed genes
# If you simply run the algorithm, all the steps will be run at once
# You can do it stepwise, which is helpful for cross-validation sometimes when you 
# want to tune parameters in later steps. You don't need to re-run earlier steps 

algorithm = Node2Vec()
algorithm.run(G=pp, seed_nodes=seed_nodes)

# Or do this stepwise with the same runtime
# Because once the walks done or embedding learnt, the algorithm just proceeds to next step

algorithm.simulate_walks(G=pp)
algorithm.learn_embeddings()
algorithm.run(G=pp, seed_nodes=seed_nodes)

# Getting results is the same for all algorithms
print(algorithm.results)
print(algorithm.get_results_df())

import sys
import os
import networkx as nx
import pandas as pd

sys.path.extend(os.getcwd())
from gp_tools.methods.randomwalk import RandomWalk
from gp_tools.methods.cv_tuning import LOOCV

# import a protein-protein interaction graph
pp = nx.read_edgelist('./project_data/graphs/ppiall.edgelist')

# Seed genes
seed_genes = pd.read_csv('./project_data/data/seed_genes.tsv', sep='\t')
seed_nodes = list(seed_genes.seeds)

# Algorithm
algorithm = RandomWalk()

loocv = LOOCV(base_algorithm=algorithm) # n_jobs here controls threads used for paralel computing
loocv.cv(G=pp, seed_nodes=seed_nodes)

print(loocv.results)
print(loocv.get_results_df())

import sys
import os
import networkx as nx
import pandas as pd

sys.path.extend(os.getcwd())
from gp_tools.methods.randomwalk import RandomWalk
from gp_tools.methods.cv_tuning import LOOCV
from gp_tools.methods.cv_tuning import GridTuneCV

# import a protein-protein interaction graph
pp = nx.read_edgelist('./project_data/graphs/ppiall.edgelist')

# Seed genes
seed_genes = pd.read_csv('./project_data/data/seed_genes.tsv', sep='\t')
seed_nodes = list(seed_genes.seeds)

# Algorithm
algorithm = RandomWalk()
loocv = LOOCV(base_algorithm=algorithm) # n_jobs here controls threads used for paralel computing

params_to_tune = {
    'gamma': [0.1, 0.5, 0.9],
    'tol': 1e-08,
    'max_iter': [10, 100],
    'protein_identifier': 'ENS'
}

tune = GridTuneCV(base_cv=loocv, params_to_tune=params_to_tune)
tune.tune(G=pp, seed_nodes=seed_nodes)
print(tune.get_results_df())