

In [1]:

    
import numpy as np



In [2]:

    
import os



In [3]:

    
import pandas as pd



In [4]:

    
from sklearn.cluster import bicluster

2015 December 4-6

Loading and exploring UTSW RNAi dataset...



In [5]:

    
dfFile = os.path.join('..', 'data', 'siRNA_dataframe.csv')



In [6]:

    
RNAiDf = pd.read_csv(dfFile, index_col=0)



In [9]:

    
RNAiDf.tail()

Impute NaN values by replacing with the column median. Choosing the column b/c more points exist to give better statistics along with the attempting to establish that the gene deletion in question has little effect on the cell line.



In [7]:

    
rplVals = np.nanmedian(RNAiDf.values, axis=0)



In [8]:

    
for i,col in enumerate(RNAiDf.columns):
    RNAiDf[col].replace(np.nan, rplVals[i], inplace=True)

Spectral Co-Clustering



In [9]:

    
modelSC = bicluster.SpectralCoclustering()



In [10]:

    
modelSC.fit(RNAiDf.values)

Spectral Biclustering



In [11]:

    
modelSB = bicluster.SpectralBiclustering()



In [12]:

    
modelSB.fit(RNAiDf.values)



In [ ]:

	HBEC30	H1155	HCC366	H1819	HCC44	HCC4017	H1993	H460	H2073	H2009	H2122	H1395	HCC95
54553	NaN	0.789991	-0.520375	1.868076	0.383222	-0.741300	0.027237	1.527195	0.151016	-0.070423	0.092664	0.280136	0.91
54729	NaN	-1.022032	-3.813564	0.266868	-7.329748	-0.474432	-8.524133	-2.004592	-2.934604	-1.527374	-0.520543	-0.540677	-0.70
54753	NaN	-0.363565	-0.320215	1.334340	0.652826	1.037820	-0.397525	1.052426	-0.236109	-0.043509	0.639026	-0.181569	0.76
54906	NaN	0.485187	-0.234777	0.118608	0.152439	1.230558	0.271865	0.848898	1.423527	-0.002306	1.346933	0.865910	-0.19
54944	NaN	1.102957	0.397106	0.252042	3.988061	1.215732	3.433279	1.975440	1.119274	1.999617	0.728126	-1.167717	-0.17