Generation of a joint probability consensus matrix from pandas dataframe

jprob_cmatrix.py

import numpy as np
import pandas as pd

# load data
mat = pd.read_table('matrix.txt', index_col=0)

# get classes
classes = np.unique(mat.values)
classes = classes[~np.isnan(classes)]

# create support dataframe
sup = pd.DataFrame(index=mat.index, columns=classes)

# get class probabilities for each sample
for ind in mat.index:
    vc = mat.ix[ind].value_counts()
    for cls in vc.index:
        sup.ix[ind, cls] = float(vc.ix[cls]) / vc.sum()

sup = sup.fillna(0).astype(float)

# generate consensus matrix
cmat = sup.dot(sup.T)

# to output
cmat.to_csv('consensus_mat.txt', sep='\t', index_label='Samples')