def comparedWithRmsd(fixed_rmsd_xcms):
def calculateAUC(scores):
scores["native_like"] = scores.rmsd.apply(
lambda r: 1 if r < 3 else 0)
fpr, tpr, thresholds = metrics.roc_curve(scores.native_like,
scores.spearmanr,
pos_label=1)
return metrics.auc(fpr, tpr)
'''
# averaged spearmanr
print(
"################################################################################")
ave_spearmanr = fixed_rmsd_xcms.groupby("query").apply(
lambda g: g[["spearmanr", "rmsd"]].mean())
print("AUC:", calculateAUC(ave_spearmanr))
pprint(ave_spearmanr.corr())
pprint(pd.Series(minepy.minestats(ave_spearmanr.rmsd,
ave_spearmanr.spearmanr)))
# ave_spearmanr.plot(kind='scatter', x='rmsd', y="spearmanr")
# plt.savefig('rmsd_ave_spearmanr.png')
# tc_times_ps weighted spearmanr
def weighted(g):
score = g["tc_times_ps"].dot(g["spearmanr"]) / g[
"tc_times_ps"].sum()
return pd.DataFrame(
{"rmsd": g["rmsd"].tolist()[0],
"spearmanr": score},
index=[0])
# tc_times_ps weighted spearmanr
print(
"################################################################################")
weighted_spearmanr = fixed_rmsd_xcms.groupby("query").apply(
weighted)
print("AUC:", calculateAUC(weighted_spearmanr))
pprint(weighted_spearmanr.corr())
pprint(pd.Series(minepy.minestats(weighted_spearmanr.rmsd,
weighted_spearmanr.spearmanr)))
# weighted_spearmanr.plot(kind='scatter', x='rmsd', y="spearmanr")
# plt.savefig('rmsd_weighted_spearmanr.png')
'''
# tc_times_ps ranked spearmanr
print(
"################################################################################")
ranked_spearmanr = fixed_rmsd_xcms.groupby("query").apply(
lambda g: g.sort_values("tc_times_ps", ascending=False).iloc[0][["spearmanr", "rmsd"]])
print("AUC:", calculateAUC(ranked_spearmanr))
print
pprint(ranked_spearmanr.corr())
print
pprint(pd.Series(minepy.minestats(ranked_spearmanr.rmsd,
ranked_spearmanr.spearmanr)))
def doCorrelation(vectorX, vectorY):
corrList_=[]
#from minepy import minepy
from scipy import stats
import minepy
## do not chnage alpha and c in the following line
minedValue = minepy.minestats(vectorX, vectorY, alpha=0.6, c=15)
spearmanValue = stats.spearmanr(vectorX, vectorY)
pearsonValue = stats.pearsonr(vectorX, vectorY)
corrList_ =[pearsonValue[0] , pearsonValue[1], spearmanValue[0], spearmanValue[1], minedValue['mic'], minedValue['nonlinearity']]
return corrList_
def doCorrelation(vectorX, vectorY):
corrList_ = []
#from minepy import minepy
from scipy import stats
import minepy
## do not chnage alpha and c in the following line
minedValue = minepy.minestats(vectorX, vectorY, alpha=0.6, c=15)
spearmanValue = stats.spearmanr(vectorX, vectorY)
pearsonValue = stats.pearsonr(vectorX, vectorY)
corrList_ = [
pearsonValue[0], pearsonValue[1], spearmanValue[0], spearmanValue[1],
minedValue['mic'], minedValue['nonlinearity']
]
return corrList_
def performMIC(combined_list, l1, l2):
print (l1, l2, len(combined_list))
outfile2 = open(prots+"_mic"+".csv", 'w')
outfile2.writelines('first,' + 'second,' + 'mic,' + '\n')
for counter1 in range(0, l1-1):
for counter2 in range(l1, len(combined_list)-1):
mine_result = minestats(combined_list[counter1],combined_list[counter2],alpha=0.6, c=15)
print (mine_result.items())
outfile2.writelines(str(counter1)+','+str(counter2-l1)+','+str(mine_result['mic']))
outfile2.write('\n')
outfile2.flush()
outfile2.close()
def correlation(df):
tc_times_ps = df.Tc * df.ps_score
print("Maximal information-based stats")
pprint(minepy.minestats(tc_times_ps, df['spearmanr']))
sph_obs = int(np.round(1/(tsol[1]-tsol[0])))
ts_data = sol[:,[0,2,4,6]]
ts_pre_fix = np.copy(ts_data)
ts_data = np.zeros([np.floor(sph*len(ts_data)/sph_obs),cellcount])
for i in range(len(ts_data)):
ts_data[i,:] = np.sum(ts_pre_fix[
int(i*sph_obs/sph):int((1+i)*sph_obs/sph),:],axis=0)
mic = np.zeros([cellcount**2,4])
aa = [range(cellcount),range(cellcount)]
inds = list(itertools.product(*aa))
for i in xrange(len(inds)):
[c1,c2] = inds[i]
ts1 = ts_data[:,c1]
ts2 = ts_data[:,c2]
mic[i,:] = [c1, c2, mp.minestats(ts1,ts2)['mic'],
stats.linregress(ts1,ts2)[2]]
hmp = np.zeros([cellcount,cellcount])
hmp2 = np.copy(hmp)
for i in range(len(mic)):
ind1 = mic[i][0]
ind2 = mic[i][1]
hmp[ind1,ind2] = mic[i][2]
hmp2[ind1,ind2] = mic[i][3]
np.fill_diagonal(hmp,0)
plt.figure(2)
plt.pcolormesh(hmp2)
plt.colorbar()