if not args.train_str_model:
check_cmd += " -onlyseq"
print(check_cmd)
output = subprocess.getoutput(check_cmd)
params_file = args.data_id + ".params"
assert os.path.exists(
params_file
), "Hyperparameter optimization output " ' .params file "%s" not found' % (
params_file)
# Add model type to params file.
if args.train_str_model:
gplib.echo_add_to_file("model_type: structure", params_file)
else:
gplib.echo_add_to_file("model_type: sequence", params_file)
# Get parameter string.
param_string = gplib.graphprot_get_param_string(params_file)
"""
Do the model training. (Yowza!)
"""
print("Starting model training (-action train) ... ")
check_cmd = ("GraphProt.pl -action train -prefix " + args.data_id +
" -fasta " + pos_train_fa + " -negfasta " + neg_train_fa +
" " + param_string)
print(check_cmd)
output = subprocess.getoutput(check_cmd)
assert output, "The following call of GraphProt.pl produced no output:\n%s" % (
check_cmd)
if args.gp_output:
print(output)
model_file = args.data_id + ".model"
assert (seq_id in reg_len_dic
), 'sequence ID "%s" missing in input .bed "%s"' % (
seq_id,
args.genomic_sites_bed,
)
reg_l = reg_len_dic[seq_id]
assert (
seq_l == reg_l
), "sequence length differs from .bed region length (%i != %i)" % (
seq_l,
reg_l,
)
# Read in model parameters.
param_dic = gplib.graphprot_get_param_dic(args.in_params)
# Create GraphProt parameter string.
param_string = gplib.graphprot_get_param_string(args.in_params)
"""
Run predictions.
"""
if args.ws_pred:
# Do whole site prediction.
print("Starting whole site predictions on input .fa file"
" (-action predict) ... ")
check_cmd = ("GraphProt.pl -action predict -prefix " + args.data_id +
" -fasta " + args.in_fa + " " + param_string +
" -model " + args.in_model)
output = subprocess.getoutput(check_cmd)
assert (
output
), "the following call of GraphProt.pl " "produced no output:\n%s" % (