

In [1]:

    
from cno.io import xcnograph
%pylab inline
matplotlib.rcParams['figure.figsize'] = (12,8)









    



Couldn't import dot_parser, loading of dot files will not be possible.
Populating the interactive namespace from numpy and matplotlib



In [2]:

    
c = xcnograph.XCNOGraph()

random graph

Let us create a random graph that 'looks like' typical graph used in CellNOpt: some ligands, some readouts (transcripts without output:

5 ligands (green); in degree is zero
14 measured (blue) with 5 transcripts (out degree set to 0)
10 other nodes (white)

Note that there are few inhibitors (10% of the edges) and no self loops, which is optional



In [3]:

    
c.random_cnograph()



In [4]:

    
c.png









    Out[4]:

double edges swap

two edges are selected randomly:

u v            u v
| |  becomes   | |
x y            y x

Their edges are swapped.

If u-y or v-x already exists, no swap is performed
and gates are ignored (removed at the beginning). They can then be added back by the user.
If the network becomes unconnected, the proposed swap is skipped.
If a self loop is created, the proposed swap is skipped
If the indegree or outdegree changes, the swap is skipped.



In [5]:

    
res = c.swap_edges(100)



In [6]:

    
c.png









    Out[6]:

You can figure out the similarity between the orignal and new graph using for isntance the number of common edges as a measure of similarity.



In [7]:

    
from cno.io import randomisation
reload(randomisation)









    Out[7]:





<module 'cno.io.randomisation' from '/home/cokelaer/Work/github/cellnopt/cno/io/randomisation.pyc'>



In [8]:

    
r = randomisation.RandomGraph()
r.parameters









    Out[8]:





{'extraNode': 10, 'nSignals': 4, 'nStimuli': 4, 'nTrans': 4}



In [9]:

    
# We will (1) create a random graph with default number of nodes
# (2) swap edges 150 times and (3) repeat this 10 times
r.run(100, 100, verbose=False)









    



fixing stimulus L1






    



---------------------------------------------------------------------------
KeyError                                  Traceback (most recent call last)
<ipython-input-9-042761c21f44> in <module>()
      1 # We will (1) create a random graph with default number of nodes
      2 # (2) swap edges 150 times and (3) repeat this 10 times
----> 3 r.run(100, 100, verbose=False)

/home/cokelaer/Work/github/cellnopt/cno/io/randomisation.pyc in run(self, N, Nswaps, verbose)
     67             if verbose:
     68                 print("Creating graph %s " % i)
---> 69             self.create_graph()
     70             if verbose:
     71                 print('Now the swapping')

/home/cokelaer/Work/github/cellnopt/cno/io/randomisation.pyc in create_graph(self, fraction)
     22                 fraction_activation=fraction, nTranscript=self.parameters['nTrans'],
     23                 nExtraNode=self.parameters['extraNode'])
---> 24         self.Nneg = [x[2]['link'] for x in self.c.edges(data=True)].count('-')
     25         self.indeg = self.c.in_degree().copy()
     26         self.outdeg = self.c.out_degree().copy()

KeyError: 'link'



In [10]:

    
r.parameters['nSignals'] = 25
r.parameters['nTrans'] = 8
r.parameters['nStimuli'] = 10
r.parameters['extraNode'] = 80



In [11]:

    
r.create_graph()



In [12]:

    
print r.c









    



The model contains 115 nodes (and 0 AND node)
344 Hyperedges found (344+0)



In [13]:

    
r.run(3, 600 , verbose=True)









    



Creating graph 0 
Now the swapping
Creating graph 1 
fixing stimulus L4
Now the swapping
Creating graph 2 
fixing stimulus L5
Now the swapping
Creating graph 3 
Now the swapping
Creating graph 4 
fixing stimulus L5
Now the swapping
Creating graph 5 
Now the swapping
Creating graph 6 
fixing stimulus L2
Now the swapping
Creating graph 7 
Now the swapping
Creating graph 8 
Now the swapping
Creating graph 9 
fixing stimulus L6
Now the swapping

conclusions

can create 'random' cno-graph
can swap edges keeping order degrees constant
can swap edges keeping ratio inhibitor/active linkes constant
slowish (2-4 seconds to swap 600 times edges on a graph of 400 edges) but seems to work
plateau is of course function of the number of edges but probably also on number of stimuli/transcript



In [ ]: