def compute_statistics(self):
"""Returns final results of the user predcition
:return: a dataframe with various metrics for each transcription factor.
Must call :meth:`score` before.
"""
data = {
'Pearson': [],
'Spearman': [],
'Pearson_Log': [],
"AUROC_8mer": [],
"AUPR_8mer": [],
"AUROC_probe": [],
"AUPR_probe": []
}
pb = progress_bar(self.Ntf, interval=1)
for tf_index in range(1, self.Ntf + 1):
dfdata = pd.read_csv(self._setfile(tf_index, "Data"),
sep='\t',
header=None)
pearson = dfdata.corr('pearson').ix[0, 1]
spearman = dfdata.corr('spearman').ix[0, 1]
pearsonLog = np.log10(dfdata).corr('pearson').ix[0, 1]
data['Pearson'].append(pearson)
data['Pearson_Log'].append(pearsonLog)
data['Spearman'].append(spearman)
dvdata = self._dvs[tf_index]
r = ROCDiscovery(dvdata.values)
rocdata = r.get_statistics()
auroc = r.compute_auc(roc=rocdata)
aupr = r.compute_aupr(roc=rocdata)
data['AUROC_8mer'].append(auroc)
data['AUPR_8mer'].append(aupr)
dvdata = self._dvps[tf_index]
r = ROCDiscovery(dvdata.values)
rocdata = r.get_statistics()
auroc = r.compute_auc(roc=rocdata)
aupr = r.compute_aupr(roc=rocdata)
data['AUROC_probe'].append(auroc)
data['AUPR_probe'].append(aupr)
pb.animate(tf_index)
df = pd.DataFrame(data)
df = df[[
'Pearson', u'Spearman', u'Pearson_Log', u'AUROC_8mer',
u'AUPR_8mer', u'AUROC_probe', u'AUPR_probe'
]]
return df
def compute_statistics(self):
"""Returns final results of the user predcition
:return: a dataframe with various metrics for each transcription factor.
Must call :meth:`score` before.
"""
data = {'Pearson': [],
'Spearman': [],
'Pearson_Log': [],
"AUROC_8mer": [],
"AUPR_8mer": [],
"AUROC_probe": [],
"AUPR_probe": []}
pb = progress_bar(self.Ntf, interval=1)
for tf_index in range(1, self.Ntf + 1):
dfdata = pd.read_csv(self._setfile(tf_index, "Data"), sep='\t', header=None)
pearson = dfdata.corr('pearson').ix[0,1]
spearman = dfdata.corr('spearman').ix[0,1]
pearsonLog = np.log10(dfdata).corr('pearson').ix[0,1]
data['Pearson'].append(pearson)
data['Pearson_Log'].append(pearsonLog)
data['Spearman'].append(spearman)
dvdata = self._dvs[tf_index]
r = ROCDiscovery(dvdata.values)
rocdata = r.get_statistics()
auroc = r.compute_auc(roc=rocdata)
aupr = r.compute_aupr(roc=rocdata)
data['AUROC_8mer'].append(auroc)
data['AUPR_8mer'].append(aupr)
dvdata = self._dvps[tf_index]
r = ROCDiscovery(dvdata.values)
rocdata = r.get_statistics()
auroc = r.compute_auc(roc=rocdata)
aupr = r.compute_aupr(roc=rocdata)
data['AUROC_probe'].append(auroc)
data['AUPR_probe'].append(aupr)
pb.animate(tf_index)
df = pd.DataFrame(data)
df = df[['Pearson', u'Spearman', u'Pearson_Log', u'AUROC_8mer',
u'AUPR_8mer', u'AUROC_probe', u'AUPR_probe']]
return df
def score(self, filename):
"""
:return: dictionay with AUC/AUPR metrics and score.
"""
self._load_proba()
prediction = pd.read_csv(filename, sep='[ \t]', engine='python', header=None)
gold = pd.read_csv(self.download_goldstandard(), sep='[ \t]',
engine='python', header=None)
prediction.columns = ['sequence', 'value']
gold.columns = ['sequence', 'value']
# merge the prediction and gold based on the sequence.
data = pd.merge(prediction, gold, how='inner', on=['sequence'],
suffixes=['_pred', '_gold'])
# sory by prediction
try:
data.sort_values(by=['value_pred'], ascending=False, inplace=True)
except:
data.sort(columns=['value_pred'], ascending=False, inplace=True)
data.columns = ['Sequence', 'prediction_values', 'prediction']
self.data = data
from dreamtools.core.rocs import ROCDiscovery
self.roc = ROCDiscovery(self.data['prediction'])
self.roc.get_statistics()
auroc = self.roc.compute_auc()
aupr = self.roc.compute_aupr()
P_AUPR = self._probability(self.aupr['X'][0], self.aupr['Y'][0], aupr)
P_AUROC = self._probability(self.auroc['X'][0], self.auroc['Y'][0], auroc)
score = np.mean(-np.log10([P_AUROC, P_AUPR]))
return {'auroc':auroc, 'aupr':aupr, 'pval_aupr': P_AUPR,
'pval_auroc':P_AUROC, 'score':score}
def score(self, filename):
"""
:return: dictionay with AUC/AUPR metrics and score.
"""
self._load_proba()
prediction = pd.read_csv(filename, sep='[ \t]', engine='python', header=None)
gold = pd.read_csv(self.download_goldstandard(), sep='[ \t]',
engine='python', header=None)
prediction.columns = ['sequence', 'value']
gold.columns = ['sequence', 'value']
# merge the prediction and gold based on the sequence.
data = pd.merge(prediction, gold, how='inner', on=['sequence'],
suffixes=['_pred', '_gold'])
# sory by prediction
try:
data.sort_values(by=['value_pred'], ascending=False, inplace=True)
except:
data.sort(columns=['value_pred'], ascending=False, inplace=True)
data.columns = ['Sequence', 'prediction_values', 'prediction']
self.data = data
from dreamtools.core.rocs import ROCDiscovery
self.roc = ROCDiscovery(self.data['prediction'])
self.roc.get_statistics()
auroc = self.roc.compute_auc()
aupr = self.roc.compute_aupr()
P_AUPR = self._probability(self.aupr['X'][0], self.aupr['Y'][0], aupr)
P_AUROC = self._probability(self.auroc['X'][0], self.auroc['Y'][0], auroc)
score = np.mean(-np.log10([P_AUROC, P_AUPR]))
return {'auroc':auroc, 'aupr':aupr, 'pval_aupr': P_AUPR,
'pval_auroc':P_AUROC, 'score':score}
class D5C1(Challenge):
"""A class dedicated to D5C1 challenge
::
from dreamtools import D5C1
s = D5C1()
filename = s.download_template()
s.score(filename)
Data and templates are downloaded from Synapse. You must have a login.
"""
def __init__(self):
""".. rubric:: constructor
"""
super(D5C1, self).__init__('D5C1')
self._path2data = os.path.split(os.path.abspath(__file__))[0]
self._init()
self.sub_challenges = []
def _init(self):
# should download files from synapse if required.
self._download_data('AUPR.mat', 'syn4560154')
self._download_data('AUROC.mat', 'syn4560158')
self._download_data('DREAM5_EAR_GoldStandard.tsv', 'syn4560182')
self._download_data('DREAM5_EAR_myteam_Predictions.txt', 'syn4560167')
def download_template(self):
# should return full path to a template file
return self.get_pathname('DREAM5_EAR_myteam_Predictions.txt')
def download_goldstandard(self):
# should return full path to a gold standard file
return self.get_pathname('DREAM5_EAR_GoldStandard.tsv')
def _load_proba(self):
import scipy.io
self.auroc = scipy.io.loadmat(self.get_pathname("AUROC.mat"))
self.aupr = scipy.io.loadmat(self.get_pathname("AUPR.mat"))
def score(self, filename):
self._load_proba()
prediction = pd.read_csv(filename, sep='[ \t]', engine='python', header=None)
gold = pd.read_csv(self.download_goldstandard(), sep='[ \t]',
engine='python', header=None)
prediction.columns = ['sequence', 'value']
gold.columns = ['sequence', 'value']
# merge the prediction and gold based on the sequence.
data = pd.merge(prediction, gold, how='inner', on=['sequence'],
suffixes=['_pred', '_gold'])
# sory by prediction
data.sort(columns=['value_pred'], ascending=False, inplace=True)
data.columns = ['Sequence', 'prediction_values', 'prediction']
self.data = data
from dreamtools.core.rocs import ROCDiscovery
self.roc = ROCDiscovery(self.data['prediction'])
self.roc.get_statistics()
auroc = self.roc.compute_auc()
aupr = self.roc.compute_aupr()
P_AUPR = self._probability(self.aupr['X'][0], self.aupr['Y'][0], aupr)
P_AUROC = self._probability(self.auroc['X'][0], self.auroc['Y'][0], auroc)
# overall dream score
#i#P = [ p_auroc p_aupr ];
#o#verall_score = mean(-log10(P)')';
score = np.mean(-np.log10([P_AUROC, P_AUPR]))
return {'auroc':auroc, 'aupr':aupr, 'pval_aupr': P_AUPR, 'pval_auroc':P_AUROC,
'score':score}
def _probability(self, X, Y, x):
dx = X[2] - X[1]
return sum( Y[X>=x] * dx )
class D5C1(Challenge):
"""A class dedicated to D5C1 challenge
::
from dreamtools import D5C1
s = D5C1()
filename = s.download_template()
s.score(filename)
"""
def __init__(self, verbose=True, download=True, **kargs):
""".. rubric:: constructor"""
super(D5C1, self).__init__('D5C1', verbose, download)
self._init()
self.sub_challenges = []
def _init(self):
if self._standalone is True:
return
# should download files from synapse if required.
self._download_data('AUPR.mat', 'syn4560154')
self._download_data('AUROC.mat', 'syn4560158')
self._download_data('DREAM5_EAR_GoldStandard.tsv', 'syn4560182')
self._download_data('DREAM5_EAR_myteam_Predictions.txt', 'syn4560167')
def download_template(self):
# should return full path to a template file
return self.get_pathname('DREAM5_EAR_myteam_Predictions.txt')
def download_goldstandard(self):
# should return full path to a gold standard file
return self.get_pathname('DREAM5_EAR_GoldStandard.tsv')
def _load_proba(self):
import scipy.io
self.auroc = scipy.io.loadmat(self.get_pathname("AUROC.mat"))
self.aupr = scipy.io.loadmat(self.get_pathname("AUPR.mat"))
def score(self, filename):
"""
:return: dictionay with AUC/AUPR metrics and score.
"""
self._load_proba()
prediction = pd.read_csv(filename, sep='[ \t]', engine='python', header=None)
gold = pd.read_csv(self.download_goldstandard(), sep='[ \t]',
engine='python', header=None)
prediction.columns = ['sequence', 'value']
gold.columns = ['sequence', 'value']
# merge the prediction and gold based on the sequence.
data = pd.merge(prediction, gold, how='inner', on=['sequence'],
suffixes=['_pred', '_gold'])
# sory by prediction
try:
data.sort_values(by=['value_pred'], ascending=False, inplace=True)
except:
data.sort(columns=['value_pred'], ascending=False, inplace=True)
data.columns = ['Sequence', 'prediction_values', 'prediction']
self.data = data
from dreamtools.core.rocs import ROCDiscovery
self.roc = ROCDiscovery(self.data['prediction'])
self.roc.get_statistics()
auroc = self.roc.compute_auc()
aupr = self.roc.compute_aupr()
P_AUPR = self._probability(self.aupr['X'][0], self.aupr['Y'][0], aupr)
P_AUROC = self._probability(self.auroc['X'][0], self.auroc['Y'][0], auroc)
score = np.mean(-np.log10([P_AUROC, P_AUPR]))
return {'auroc':auroc, 'aupr':aupr, 'pval_aupr': P_AUPR,
'pval_auroc':P_AUROC, 'score':score}
def _probability(self, X, Y, x):
dx = X[2] - X[1]
return sum( Y[X>=x] * dx )