In [1]:

    

%pylab inline


    

    




Populating the interactive namespace from numpy and matplotlib

In [9]:

    

import pulp
from cno import cnodata, CNOGraph, XMIDAS


    

In [10]:

    

import cno.milp.model


    

In [82]:

    

# select a prior knowledge network (pkn) and a dataset in midas format
filename_pkn = cnodata("PKN-ToyMMB.sif")
filename_midas = cnodata("MD-ToyMMB.csv")

#filename_pkn = cnodata("PKN-LiverDREAM.sif")
#filename_midas = cnodata("MD-LiverDREAM.csv")


    

In [83]:

    

# load pkn and midas
pkn = CNOGraph(filename_pkn, filename_midas)
midas = pkn.midas


    

In [84]:

    

# we can compress the network to reduce the problem size
pkn.compress()


    

In [85]:

    

pkn.expand_and_gates()


    

In [86]:

    

# instantiate a MILPTrain object from a given pkn and midas data and solve the optimization problem
milp_model = cno.milp.model.MILPTrain(pkn, midas)
milp_model.train()
print('Error:', pulp.value(milp_model.error_objective_expression()))
print('Size:', pulp.value(milp_model.size_objective_expression()))


    

    




('Error:', 10.01999999999998)
('Size:', 9.0)

In [87]:

    

# we can retrieve a dictionary with the reactions as keys and a 1 or a 0 as value 
# indicating if the reaction is present in the optimized network or not
rxn_sol = milp_model.get_rxn_solution()


    

In [88]:

    

pkn2 = CNOGraph()


    

In [89]:

    

for i in rxn_sol:
    if rxn_sol[i]==1:
        pkn2.add_reaction(i)


    

In [90]:

    

pkn2.add_nodes_from(pkn.species)
pkn2.midas = midas
pkn2.png


    

    
Out[90]:

In [91]:

    

# solution from GA cellnopt
bs = [1, 1, 0, 1, 1, 0, 1, 1, 0, 0, 1, 1, 1, 0, 0, 0]
reactions = ['EGF=PI3K', 'TNFa=PI3K', 'Jnk=cJun', 'PI3K=Akt', 'Raf=Mek',
        '!Akt=Mek', 'Mek=p90RSK', 'Mek=Erk', 'Erk=Hsp27', 'TNFa=Jnk',
        'TNFa=NFkB', 'TNFa=Hsp27', 'EGF=Raf', 'EGF+TNFa=PI3K',
        'Raf+!Akt=Mek', 'Erk+TNFa=Hsp27']


    

In [92]:

    

pkn3 = CNOGraph()
for b, r in zip(bs, reactions):
    if b==1: pkn3.add_reaction(r.replace("+","^"))
pkn3.add_nodes_from(pkn.species)
pkn3.midas = midas


    

In [93]:

    

pkn2 == pkn3


    

    
Out[93]:





True

In [94]:

    

pkn3.png


    

    
Out[94]:

In [95]:

    

pkn2.png


    

    
Out[95]:

In [ ]: