preds = self.clf[it].predict(itinsts[columns].values)
itinsts["ptag"] = preds
cprobs = self.clf[it].predict_proba(itinsts[columns].values)
tpcol = np.where(self.clf[it].classes_ == "TP")[0][0]
itinsts["qual"] = cprobs[:, tpcol]
insts.append(itinsts)
return pandas.concat(insts)
def plots(self, prefix, featurenames):
""" Make diagnostic plots """
for it in self.itypes:
importances = self.clf[it].feature_importances_
std = np.std([
tree.feature_importances_ for tree in self.clf[it].estimators_
],
axis=0)
indices = np.argsort(importances)[::-1]
# Print the feature ranking
print "Feature ranking for INDELTYPE == %s:" % str(it)
for f in xrange(0, len(indices)):
print "%d. feature %d:%s (%f +- %f)" % (
f + 1, indices[f], featurenames[indices[f]],
importances[indices[f]], std[indices[f]])
EVSModel.register("strelka.rf.indel", StrelkaRFIndel)
axis=0)
indices = np.argsort(importances)[::-1]
# Print the feature ranking
print "Feature ranking:"
for f in xrange(0, len(indices)):
print "%d. feature %d:%s (%f +- %f)" % (
f + 1, indices[f], featurenames[indices[f]],
importances[indices[f]], std[indices[f]])
def draw_trees(self):
""" Draw trees in png files """
from sklearn.tree import export_graphviz
import subprocess
images = []
for i, tree in enumerate(self.clf.estimators_):
# creates indel_tree_...dot files
fname = 'trees/tree_' + str(i) + '.dot'
with open(fname, 'w') as dotfile:
export_graphviz(tree, dotfile, feature_names=featurenames)
# run dot
subprocess.check_call(
["dot -Tpng " + fname + " > " + fname + ".png"], shell=True)
images.append(Image(fname + ".png"))
EVSModel.register("germline.rf", GermlineRF)
all_trees = io.classifier_to_dict(self.clf)
full_model = meta
full_model["Model"] = all_trees
modelFile = {
"CalibrationModels": {
"Somatic": {
varianttype: full_model
}
}
}
json.dump(modelFile, open(filename, "wb"))
def plots(self, prefix, featurenames):
""" Make diagnostic plots """
importances = self.clf.feature_importances_
std = np.std(
[tree.feature_importances_ for tree in self.clf.estimators_],
axis=0)
indices = np.argsort(importances)[::-1]
# Print the feature ranking
print "Feature ranking:"
for f in xrange(0, len(indices)):
print "%d. feature %d:%s (%f +- %f)" % (
f + 1, indices[f], featurenames[indices[f]],
importances[indices[f]], std[indices[f]])
EVSModel.register("somatic.rf", SomaticRF)
importances = self.clf.feature_importances_
std = np.std(
[tree.feature_importances_ for tree in self.clf.estimators_],
axis=0)
indices = np.argsort(importances)[::-1]
# Print the feature ranking
print "Feature ranking:"
for f in xrange(0, len(indices)):
print "%d. feature %d:%s (%f +- %f)" % (
f + 1, indices[f], featurenames[indices[f]],
importances[indices[f]], std[indices[f]])
# images = []
# for i, tree in enumerate(self.clf.estimators_):
# creates indel_tree_...dot files
# fname = 'trees112/tree_' + str(i) + '.dot'
# with open(fname, 'w') as dotfile:
# export_graphviz(
# tree,
# dotfile,
# feature_names=featurenames)
# run dot
# subprocess.check_call(["dot -Tpng " + fname + " > " + fname + ".png"], shell=True)
# images.append(Image(fname + ".png"))
EVSModel.register("strelka.rf", StrelkaRF)