def tool_newref(args):
logging.info('Creating new reference')
split_path = list(os.path.split(args.outfile))
if split_path[-1][-4:] == '.npz':
split_path[-1] = split_path[-1][:-4]
base_path = os.path.join(split_path[0], split_path[1])
args.basepath = base_path
args.prepfile = '{}_prep.npz'.format(base_path)
args.partfile = '{}_part'.format(base_path)
samples = []
logging.info('Importing data ...')
for infile in args.infiles:
logging.info('Loading: {}'.format(infile))
npzdata = np.load(infile, encoding='latin1')
sample = npzdata['sample'].item()
binsize = int(npzdata['binsize'])
logging.info('Binsize: {}'.format(int(binsize)))
samples.append(scale_sample(sample, binsize, args.binsize))
samples = np.array(samples)
genders, trained_cutoff = train_gender_model(samples)
if not args.nipt:
for i, sample in enumerate(samples):
samples[i] = gender_correct(sample, genders[i])
total_mask, bins_per_chr = get_mask(samples)
if genders.count('F') > 4:
mask_F, _ = get_mask(samples[np.array(genders) == 'F'])
total_mask = total_mask & mask_F
if genders.count('M') > 4 and not args.nipt:
mask_M, _ = get_mask(samples[np.array(genders) == 'M'])
total_mask = total_mask & mask_M
outfiles = []
if len(genders) > 9:
logging.info('Starting autosomal reference creation ...')
args.tmpoutfile = '{}.tmp.A.npz'.format(args.basepath)
outfiles.append(args.tmpoutfile)
tool_newref_prep(args, samples, 'A', total_mask, bins_per_chr)
logging.info('This might take a while ...')
tool_newref_main(args, args.cpus)
else:
logging.critical(
'Provide at least 10 samples to enable the generation of a reference.'
)
sys.exit()
if genders.count('F') > 4:
logging.info('Starting female gonosomal reference creation ...')
args.tmpoutfile = '{}.tmp.F.npz'.format(args.basepath)
outfiles.append(args.tmpoutfile)
tool_newref_prep(args, samples[np.array(genders) == 'F'], 'F',
total_mask, bins_per_chr)
logging.info('This might take a while ...')
tool_newref_main(args, 1)
else:
logging.warning(
'Provide at least 5 female samples to enable normalization of female gonosomes.'
)
if not args.nipt:
if genders.count('M') > 4:
logging.info('Starting male gonosomal reference creation ...')
args.tmpoutfile = '{}.tmp.M.npz'.format(args.basepath)
outfiles.append(args.tmpoutfile)
tool_newref_prep(args, samples[np.array(genders) == 'M'], 'M',
total_mask, bins_per_chr)
tool_newref_main(args, 1)
else:
logging.warning(
'Provide at least 5 male samples to enable normalization of male gonosomes.'
)
tool_newref_merge(args, outfiles, trained_cutoff)
logging.info('Finished creating reference')
def tool_test(args):
logging.info('Starting CNA prediction')
if not args.bed and not args.plot:
logging.critical(
'No output format selected. '
'Select at least one of the supported output formats (--bed, --plot)'
)
sys.exit()
if args.zscore <= 0:
logging.critical(
'Parameter --zscore should be a strictly positive number')
sys.exit()
if args.beta is not None:
if args.beta <= 0 or args.beta > 1:
logging.critical(
'Parameter --beta should be a strictly positive number lower than 1'
)
sys.exit()
if args.alpha <= 0 or args.alpha > 1:
logging.critical(
'Parameter --alpha should be a strictly positive number lower than 1'
)
sys.exit()
logging.info('Importing data ...')
ref_file = np.load(args.reference, encoding='latin1')
sample_file = np.load(args.infile, encoding='latin1')
sample = sample_file['sample'].item()
n_reads = sum([sum(sample[x]) for x in sample.keys()])
sample = scale_sample(sample, int(sample_file['binsize'].item()),
int(ref_file['binsize']))
if not ref_file['is_nipt']:
actual_gender = predict_gender(sample, ref_file['trained_cutoff'])
if args.gender:
actual_gender = args.gender
sample = gender_correct(sample, actual_gender)
else:
actual_gender = 'F'
if args.gender:
actual_gender = args.gender
ref_gender = actual_gender
logging.info('Normalizing autosomes ...')
results_r, results_z, results_w, ref_sizes, m_lr, m_z = normalize(
args, sample, ref_file, 'A')
if not ref_file['has_male'] and actual_gender == 'M':
logging.warning(
'This sample is male, whilst the reference is created with fewer than 5 males. '
'The female gonosomal reference will be used for X predictions. Note that these might '
'not be accurate. If the latter is desired, create a new reference and include more '
'male samples.')
ref_gender = 'F'
elif not ref_file['has_female'] and actual_gender == 'F':
logging.warning(
'This sample is female, whilst the reference is created with fewer than 5 females. '
'The male gonosomal reference will be used for XY predictions. Note that these might '
'not be accurate. If the latter is desired, create a new reference and include more '
'female samples.')
ref_gender = 'M'
logging.info('Normalizing gonosomes ...')
null_ratios_aut_per_bin = ref_file['null_ratios']
null_ratios_gon_per_bin = ref_file['null_ratios.{}'.format(
ref_gender)][len(null_ratios_aut_per_bin):]
results_r_2, results_z_2, results_w_2, ref_sizes_2, _, _ = normalize(
args, sample, ref_file, ref_gender)
rem_input = {
'args':
args,
'wd':
str(os.path.dirname(os.path.realpath(__file__))),
'binsize':
int(ref_file['binsize']),
'n_reads':
n_reads,
'ref_gender':
ref_gender,
'actual_gender':
actual_gender,
'mask':
ref_file['mask.{}'.format(ref_gender)],
'bins_per_chr':
ref_file['bins_per_chr.{}'.format(ref_gender)],
'masked_bins_per_chr':
ref_file['masked_bins_per_chr.{}'.format(ref_gender)],
'masked_bins_per_chr_cum':
ref_file['masked_bins_per_chr_cum.{}'.format(ref_gender)]
}
del ref_file
results_r = np.append(results_r, results_r_2)
results_z = np.append(results_z, results_z_2) - m_z
results_w = np.append(results_w * np.nanmedian(results_w_2),
results_w_2 * np.nanmedian(results_w))
results_w = results_w / np.nanmedian(results_w)
ref_sizes = np.append(ref_sizes, ref_sizes_2)
null_ratios_aut_per_sample = np.transpose(null_ratios_aut_per_bin)
part_mask = np.array([not x for x in list(np.isnan(results_r))],
dtype=bool)
null_m_lr_aut = np.array([
np.nanmedian(x[part_mask[:len(null_ratios_aut_per_bin)]])
for x in null_ratios_aut_per_sample
])
null_ratios_aut_per_bin = null_ratios_aut_per_bin - null_m_lr_aut
null_ratios = np.array([x.tolist() for x in null_ratios_aut_per_bin] +
[x.tolist() for x in null_ratios_gon_per_bin])
results = {
'results_r': results_r,
'results_z': results_z,
'results_w': results_w,
'results_nr': null_ratios
}
for result in results.keys():
results[result] = get_post_processed_result(args, results[result],
ref_sizes, rem_input)
log_trans(results, m_lr)
if args.blacklist:
logging.info('Applying blacklist ...')
apply_blacklist(rem_input, results)
logging.info('Executing circular binary segmentation ...')
results['results_c'] = exec_cbs(rem_input, results)
if args.bed:
logging.info('Writing tables ...')
generate_output_tables(rem_input, results)
if args.plot:
logging.info('Writing plots ...')
exec_write_plots(rem_input, results)
logging.info('Finished prediction')