def str_display(ls):
if not isinstance(ls, (list, tuple, np.ndarray)):
str_ls = '{0:.4f}'.format(ls)
return str_ls
str_ls = ['{0:.4f}'.format(v) for v in ls ]
str_ls2 = ' '.join(str_ls)
return str_ls2
if __name__ == "__main__":
if len(sys.argv) != 4:
print 'python CalcMCCF1.py pred_matrix_file distcb_matrix_file target'
print ' Both matrix files are text format with L lines and each line has L columns where L is the protein sequence length'
exit(-1)
predFile = sys.argv[1]
distcbFile = sys.argv[2]
target = sys.argv[3]
pred = LoadContactMatrix(predFile)
truth = LoadContactMatrix(distcbFile)
for prob in np.arange(20, 60, 2):
#print "prob=", prob
accs = CalcMCCF1(pred=pred, truth=truth, probCutoff=prob/100.)
resultStr = target + ' ' + str(pred.shape[0]) + ' cutoff=' + str(prob) + ' ' + str_display(accs)
print resultStr
predInfo = cPickle.load(fh)
pred = predInfo[3]['CbCb']
preds[protein] = pred
nativeFile = os.path.join(truthDir, protein + '.native.pkl')
with open(nativeFile, 'rb') as fh:
nativeInfo = cPickle.load(fh)
truth = nativeInfo['atomDistMatrix']['CbCb']
truths[protein] = truth
for prob in np.arange(5, 60, 1):
#print "prob=", prob
accs = []
for protein in content:
acc = CalcMCCF1(pred=preds[protein],
truth=truths[protein],
probCutoff=prob / 100.)
accs.append(acc)
avgacc = np.average(accs, axis=0)
resultStr = 'per-target avgMCCF1 at cutoff=' + str(
prob) + ': ' + str_display(avgacc)
print resultStr
lrMCC = Metrics.MCC(avgacc[1], avgacc[2], avgacc[3], avgacc[4])
lrF1, lrprecision, lrrecall = Metrics.F1(avgacc[1], avgacc[2],
avgacc[3], avgacc[4])
mrMCC = Metrics.MCC(avgacc[9], avgacc[10], avgacc[11], avgacc[12])
mrF1, mrprecision, mrrecall = Metrics.F1(avgacc[9], avgacc[10],
avgacc[11], avgacc[12])