try:
gene_dir = fu.make_output_dir(experiment_dir, gene)
tcga_data.process_data_for_gene(
gene,
classification,
gene_dir,
)
except ResultsFileExistsError:
# this happens if cross-validation for this gene has already been
# run (i.e. the results file already exists)
if io_args.verbose:
print(
'Skipping because results file exists already: gene {}'
.format(gene),
file=sys.stderr)
log_df = fu.generate_log_df(
log_columns, [gene, 'N/A', shuffle_labels, 'file_exists'])
fu.write_log_file(log_df, io_args.log_file)
continue
except KeyError:
# this can happen if the given gene isn't in the mutation data
print('Gene {} not found in mutation data, skipping'.format(
gene),
file=sys.stderr)
log_df = fu.generate_log_df(
log_columns,
[gene, 'N/A', shuffle_labels, 'gene_not_found'])
fu.write_log_file(log_df, io_args.log_file)
continue
num_feats = (model_options.subset_mad_genes +
np.count_nonzero(~tcga_data.gene_features))
cancer_type_dir = fu.make_output_dir(experiment_dir,
cancer_type)
check_file = fu.check_output_file(cancer_type_dir, cancer_type,
shuffle_labels,
model_options)
tcga_data.process_data_for_cancer_type(cancer_type,
cancer_type_dir)
except ResultsFileExistsError:
# this happens if cross-validation for this cancer type has
# already been run (i.e. the results file already exists)
if io_args.verbose:
print('Skipping because results file exists already: '
'cancer type {}'.format(cancer_type),
file=sys.stderr)
log_df = fu.generate_log_df(log_columns, [
cancer_type, model_options.training_data, shuffle_labels,
'file_exists'
])
fu.write_log_file(log_df, io_args.log_file)
continue
try:
# for now, don't standardize methylation data
standardize_columns = (model_options.training_data
in cfg.standardize_data_types)
results = run_cv_stratified(
tcga_data, 'cancer_type', cancer_type,
model_options.training_data, sample_info_df,
model_options.num_folds, True, shuffle_labels,
standardize_columns, io_args.output_preds)
# only save results if no exceptions
fu.save_results(cancer_type_dir, check_file, results,
log_df = None
gene = gene_series.gene
classification = gene_series.classification
progress.set_description('gene: {}'.format(gene))
try:
tcga_data.process_data_for_gene(gene,
classification,
experiment_dir,
filter_cancer_types=False)
except KeyError:
# this can happen if the given gene isn't in the mutation data
print('Gene {} not found in mutation data, skipping'.format(gene),
file=sys.stderr)
log_df = fu.generate_log_df(
log_columns,
[gene, model_options.training_data, 'gene_not_found'])
fu.write_log_file(log_df, io_args.log_file)
continue
try:
standardize_columns = (model_options.training_data
in cfg.standardize_data_types)
results = run_cv_stratified(tcga_data,
'gene',
gene,
model_options.training_data,
sample_info_df,
model_options.num_folds,
predictor='classify',
shuffle_labels=False,
classification,
gene_dir,
batch_correction=model_options.batch_correction,
bc_cancer_type=model_options.bc_cancer_type,
drop_target=model_options.drop_target,
only_target=model_options.only_target)
except ResultsFileExistsError:
# this happens if cross-validation for this gene has already been
# run (i.e. the results file already exists)
if io_args.verbose:
print(
'Skipping because results file exists already: gene {}'
.format(gene),
file=sys.stderr)
log_df = fu.generate_log_df(log_columns, [
gene, model_options.training_data, shuffle_labels,
'file_exists'
])
fu.write_log_file(log_df, io_args.log_file)
continue
except KeyError:
# this can happen if the given gene isn't in the mutation data
print('Gene {} not found in mutation data, skipping'.format(
gene),
file=sys.stderr)
log_df = fu.generate_log_df(log_columns, [
gene, model_options.training_data, shuffle_labels,
'gene_not_found'
])
fu.write_log_file(log_df, io_args.log_file)
continue
try:
check_file = fu.check_output_file(experiment_dir, cancer_type,
shuffle_labels,
model_options, 'survival')
tcga_data.process_survival_data(experiment_dir, cancer_type)
except ResultsFileExistsError:
# this happens if cross-validation for this cancer type has already been
# run (i.e. the results file already exists)
if io_args.verbose:
print(
'Skipping because results file exists already: cancer type {}'
.format(cancer_type),
file=sys.stderr)
log_df = fu.generate_log_df(log_columns, [
cancer_type, model_options.training_data, shuffle_labels,
'file_exists'
])
fu.write_log_file(log_df, io_args.log_file)
continue
try:
# for now, don't standardize methylation data
# also, always standardize PCA components, for all data types
standardize_columns = ((model_options.n_dim is not None)
or (model_options.training_data
in cfg.standardize_data_types))
results = run_cv_stratified(
tcga_data,
'survival',
cancer_type,
model_options.training_data,