def calculateGOseq(ontology):
organism = 'thalia'
[dag, mapping] = fr.read_ontology_from_file('files/terms.obo')
[Y, termNames, _] = fr.getDataLabels(organism, True, ontology, 8000)
go = np.eye(len(termNames))
for i, tname in enumerate(termNames):
dagInd = mapping[tname]
for pInd in dag[dagInd].parents:
if dag[pInd].ontology != ontology:
continue
parLoc = termNames.index(dag[pInd].ID)
go[i, parLoc] = 1.0
go[parLoc, i] = 1.0
#[Y, go] = removeRareTermsOnly(Y, go, label_fraction)
go = csr_matrix(go)
with open('go' + ontology + '.pkl', 'wb') as f:
pickle.dump(go, f)
return go
def getParentsCoord(termNames, ontology, dag=None, mapping=None):
if dag is None:
[dag, mapping] = fr.read_ontology_from_file('files/terms.obo')
#[dag, mapping] = fr.read_ontology_from_file('files/terms_09_17.obo')
assert ontology == 'P'
root = 'GO:0008150'
parentsCoord = dict()
for i, tname in enumerate(termNames):
dagInd = mapping[tname]
parentsCoord[i] = []
for pInd in dag[dagInd].parents:
if dag[pInd].ontology != ontology or dag[pInd].ID == root:
continue
parLoc = termNames.index(dag[pInd].ID)
parentsCoord[i].append(parLoc)
return parentsCoord
def loadData(similarity, ontology, organism):
assert ontology == 'P'
root = 'GO:0008150'
[dag, mapping] = fr.read_ontology_from_file('files/terms.obo')
[Y, termNames, geneNames] = fr.getDataLabels(organism, True, ontology,
8000)
data = loadmat('files/identities.mat')
X = data['identityMatrix']
X = np.array(X)
rootInd = termNames.index(root)
Y = np.array(Y)
assert X.shape[0] == Y.shape[0]
Y = np.delete(Y, rootInd, axis=1)
del termNames[rootInd]
[X, Y, deletedProteins] = removeEmptyProteins(X, Y)
parentsCoord = getParentsCoord(termNames, ontology, dag, mapping)
for i in sorted(deletedProteins, reverse=True):
del geneNames[i]
return [X, Y, termNames, deletedProteins, parentsCoord]
def getLCAsCAFA(termNames, ontology, dag=None, mapping=None):
if dag is None:
[dag, mapping] = fr.read_ontology_from_file('files/terms_09_17.obo')
if ontology == 'P':
root = 'GO:0008150'
elif ontology == 'F':
root = 'GO:0003674'
else:
print 'wrong ontology'
sys.exit(1)
#find all ancestors of each term
ancestors = dict()
for i, t in enumerate(termNames):
ancestors[t] = set()
for pInd in dag[mapping[t]].parents:
getAncestorsRecursively(dag, pInd, ancestors[t], ontology)
ancestors[t].remove(root)
lcas = dict()
#get common ancestors
for i, t1 in enumerate(termNames):
print i
for j, t2 in enumerate(termNames):
#print t1, t2
if i == j:
lcas[t1, t2] = t1
else:
cas = ancestors[t1].intersection(ancestors[t2])
#print cas
for t in cas:
others = cas.difference(set([t]))
if others == ancestors[t]:
lcas[t1, t2] = t
break
with open('lowestCommonAncestors.pkl', 'wb') as f:
pickle.dump(lcas, f)
with open('ancestors.pkl', 'wb') as f:
pickle.dump(ancestors, f)
def calculateGOallseq(ontology):
organism = 'thalia'
[dag, mapping] = fr.read_ontology_from_file('files/terms.obo')
[Y, termNames, _] = fr.getDataLabels(organism, True, ontology, 8000)
go = DFS(dag, termNames, mapping, ontology)
go = csr_matrix(go)
with open('go_allancestors' + ontology + '.pkl', 'wb') as f:
pickle.dump(go, f)
return go
def calculateGOseqCAFA(ontology, termNames):
organism = 'thalia'
[dag, mapping] = fr.read_ontology_from_file('files/terms_09_17.obo')
if ontology == 'P':
root = 'GO:0008150'
elif ontology == 'F':
root = 'GO:0003674'
else:
print 'wrong ontology'
sys.exit(1)
go = np.eye(len(termNames))
for i, tname in enumerate(termNames):
dagInd = mapping[tname]
for pInd in dag[dagInd].parents:
if dag[pInd].ontology != ontology or dag[pInd] == root:
continue
parLoc = termNames.index(dag[pInd].ID)
go[i, parLoc] = 1.0
go[parLoc, i] = 1.0
#[Y, go] = removeRareTermsOnly(Y, go, label_fraction)
go = csr_matrix(go)
with open('go' + ontology + '.pkl', 'wb') as f:
pickle.dump(go, f)
return go
import numpy as np
import pickle
import sys
from utilities import *
from filereader import read_ontology_from_file
from gyros import getParentsCoord, calculateIC
species = sys.argv[1]
dag, mapping = read_ontology_from_file('../data/go/go-final.obo')
print('Loading go...')
[Y, geneNames, termNames, gene2row, term2col] = goLoader(species)
cv = KFold(n_splits=5, shuffle=True, random_state=656391)
np.random.seed(1901273)
thresholds = np.linspace(0, 1.0, 21)
for fold, (train, test) in enumerate(cv.split(Y)):
print fold
termNames = np.array(termNames)
Ytrain = Y[train]
Ytest = Y[test]
nonempty = np.where(np.sum(Ytest, 0) > 0)[0]
Ytest = Ytest[:, nonempty]
Y = np.delete(Y, root, 1)
return [Y, proteins, termNames]
trainProteins = set()
leaves = set()
with open('../training_data/thalia_proteins.list') as f:
for line in f:
trainProteins.add(line.rsplit('\n')[0])
print 'Proteins in training set:', len(trainProteins)
[dag, mapping] = fr.read_ontology_from_file(
'/tudelft.net/staff-bulk/ewi/insy/DBL/smakrod/go/terms_09_17.obo')
anno = dict()
absentTerms = set()
with open('../training_data/uniprot_sprot_exp.txt') as f:
for line in f:
fields = line.split('\t')
protein = fields[0]
if protein not in trainProteins:
continue