ml_dict = dict(alpha=1.0, binarize=None, fit_prior=True)
if options.ml:
ml_dict = ml_func.readMLFile(ml_dict, read_dict, path+options.ml)
# initialize machine-learning method
ml = BernoulliNB(alpha=ml_dict['alpha'], binarize=ml_dict['binarize'], fit_prior=ml_dict['fit_prior'])
# loop over targets
for target in conf.set_data:
print target
# read in training actives and calculate fps
actives = cPickle.load(open(inpath_cmp+'ChEMBL_II/Target_no_'+str(target)+'.pkl', 'r'))
for k in actives.keys():
for i,m in enumerate(actives[k]):
fp_dict = scor.getFP(fp_build, m[1])
actives[k][i] = [str(target)+'_'+str(k)+'_A_'+str(i+1), fp_dict]
# read in test actives and calculate fps
div_actives = []
for line in gzip.open(inpath_cmp+'ChEMBL/cmp_list_ChEMBL_'+str(target)+'_actives.dat.gz', 'r'):
if line[0] != '#':
# structure of line: [external ID, internal ID, SMILES]]
line = line.rstrip().split()
fp_dict = scor.getFP(fp_build, line[2])
# store: [internal ID, dict with fps]
div_actives.append([line[1], fp_dict])
num_test_actives = conf.num_div_act - 1
# convert fps to numpy arrays
np_fps_div_act = ml_func.getNumpy(div_actives)
# loop over data-set sources
for dataset in conf.set_data.keys():
print dataset
# loop over targets
for target in conf.set_data[dataset]['ids']:
print target
# read in actives and calculate fps
actives = []
for line in gzip.open(
inpath_cmp + dataset + '/cmp_list_' + dataset + '_' +
str(target) + '_actives.dat.gz', 'r'):
if line[0] != '#':
# structure of line: [external ID, internal ID, SMILES]]
line = line.rstrip().split()
fp_dict = scor.getFP(fp_build, line[2])
# store: [internal ID, dict with fps]
actives.append([line[1], fp_dict])
num_actives = len(actives)
num_test_actives = num_actives - num_query_mols
# convert fps to numpy arrays
np_fps_act = ml_func.getNumpy(actives)
# read in decoys and calculate fps
if dataset == 'ChEMBL':
if firstchembl:
decoys = []
for line in gzip.open(
inpath_cmp + dataset + '/cmp_list_' + dataset +
'_zinc_decoys.dat.gz', 'r'):
if line[0] != '#':