def test_get_sample_cats(self):
"""Test get_sample_cats."""
pmf_in = \
{'Sample1': {'test_cat': 'cat1', 'test_corr': '1', 'test_empty': 'abc'},
'Sample2': {'test_cat': 'cat1', 'test_corr': '1', 'test_empty': 'abc'},
'Sample3': {'test_cat': 'cat2', 'test_corr': '1', 'test_empty': ''},
'Sample4': {'test_cat': 'cat2', 'test_corr': '1', 'test_empty': ''},
'Sample5': {'test_cat': 'cat3', 'test_corr': '1', 'test_empty': 'abc'},
'Sample6': {'test_cat': 'cat3', 'test_corr': '1', 'test_empty': ''}}
# test with properly formatted, ordered parsed mapping file
category = 'test_cat'
exp = {'Sample1': 'cat1',
'Sample2': 'cat1',
'Sample3': 'cat2',
'Sample4': 'cat2',
'Sample5': 'cat3',
'Sample6': 'cat3'}
obs = get_sample_cats(pmf_in, category)
self.assertEqual(exp, obs)
# test with only one category
category = 'test_corr'
exp = {'Sample1': '1',
'Sample2': '1',
'Sample3': '1',
'Sample4': '1',
'Sample5': '1',
'Sample6': '1'}
obs = get_sample_cats(pmf_in, category)
self.assertEqual(exp, obs)
# test with empty categories
category = 'test_empty'
exp = {'Sample1': 'abc',
'Sample2': 'abc',
'Sample5': 'abc'}
obs = get_sample_cats(pmf_in, category)
self.assertEqual(exp, obs)
def main():
option_parser, opts, args = parse_command_line_parameters(**script_info)
# sync the mapping file and the biom file
tmp_bt = load_table(opts.otu_table_fp)
tmp_pmf, _ = parse_mapping_file_to_dict(opts.mapping_fp)
pmf, bt, nonshared_samples = sync_biom_and_mf(tmp_pmf, tmp_bt)
# test error conditions for overlapping mf and bt
if not opts.biom_samples_are_superset:
# user indicates biom sample should be subset of mapping file samples
if any([i in nonshared_samples for i in tmp_bt.ids()]):
raise ValueError('The samples in the biom table are a superset of' +
' the samples in the mapping file. The script will abort in' +
' this case even though the calculations wouldn\'t be' +
' affected, to ensure consistency within QIIME. Pass the' +
' --biom_samples_are_superset option to disable this behavior.')
# user wants non-overlapping samples printed out
if opts.print_non_overlap:
print 'The following samples were not shared between the mapping file' +\
' and the biom file and will not be included in the analysis:\n' +\
' '.join(nonshared_samples)
# find group indices
sam_cats = get_sample_cats(pmf, opts.category)
cat_sam_groups = get_cat_sample_groups(sam_cats)
cat_sam_indices = get_sample_indices(cat_sam_groups, bt)
# sanity check to prevent inscrutable errors later
if not all([len(v) > 0 for k, v in cat_sam_indices.items()]):
raise ValueError('At least one metadata group has no samples. Check ' +
'that the mapping file has at least one sample for each value in ' +
'the passed category.')
if opts.test in TWO_GROUP_TESTS and len(cat_sam_indices) > 2:
option_parser.error('The t-test and mann_whitney_u test may ' +
'only be used when there are two sample groups. Choose another ' +
'test or another metadata category.')
# check that assumptions are met for a given test:
if opts.test == 'mann_whitney_u':
sams = reduce(lambda x, y: len(x) + len(y), cat_sam_indices.values())
if sams <= 20:
raise ValueError('The number of samples is too small to use the ' +
'Mann-Whitney-U normal approximation. Review the script ' +
'documentation.')
# check that the G-test was not selected if the table appears to be
# relative abundance
if opts.test == 'g_test':
if allclose(bt.sum(axis='sample'), 1.) or (bt.sum(axis='whole') == 1.):
raise ValueError('It appears that the biom table you have passed '
'is a relative abundance table where values i,j (obsevation i '
'count in sample j) are fractional and the sum of the columns '
'is 1.0. This will fail to work properly with the G-test. If '
'your data sums to 1 in each column but your data is not '
'relative abundance then the tests will fail anyway because '
'of the reduced number of observations.')
# run actual tests
data_feed = group_significance_row_generator(bt, cat_sam_indices)
test_stats, pvals, means = run_group_significance_test(
data_feed, opts.test,
GROUP_TEST_CHOICES, int(opts.permutations))
# calculate corrected pvals
fdr_pvals = array(benjamini_hochberg_step_down(pvals))
bon_pvals = bonferroni_correction(pvals)
# correct for cases where values above 1.0 due to correction
fdr_pvals = where(fdr_pvals > 1.0, 1.0, fdr_pvals)
bon_pvals = where(bon_pvals > 1.0, 1.0, bon_pvals)
# write output results after sorting
lines = group_significance_output_formatter(bt, test_stats, pvals,
fdr_pvals, bon_pvals, means, cat_sam_indices, md_key=opts.metadata_key)
lines = sort_by_pval(lines, ind=2)
o = open(opts.output_fp, 'w')
o.writelines('\n'.join(lines))
o.close()
def main():
option_parser, opts, args = parse_command_line_parameters(**script_info)
# sync the mapping file and the biom file
tmp_bt = load_table(opts.otu_table_fp)
tmp_pmf, _ = parse_mapping_file_to_dict(opts.mapping_fp)
pmf, bt, nonshared_samples = sync_biom_and_mf(tmp_pmf, tmp_bt)
# test error conditions for overlapping mf and bt
if not opts.biom_samples_are_superset:
# user indicates biom sample should be subset of mapping file samples
if any([i in nonshared_samples for i in tmp_bt.ids()]):
raise ValueError(
'The samples in the biom table are a superset of' +
' the samples in the mapping file. The script will abort in' +
' this case even though the calculations wouldn\'t be' +
' affected, to ensure consistency within QIIME. Pass the' +
' --biom_samples_are_superset option to disable this behavior.'
)
# user wants non-overlapping samples printed out
if opts.print_non_overlap:
print 'The following samples were not shared between the mapping file' +\
' and the biom file and will not be included in the analysis:\n' +\
' '.join(nonshared_samples)
# find group indices
sam_cats = get_sample_cats(pmf, opts.category)
cat_sam_groups = get_cat_sample_groups(sam_cats)
cat_sam_indices = get_sample_indices(cat_sam_groups, bt)
# sanity check to prevent inscrutable errors later
if not all([len(v) > 0 for k, v in cat_sam_indices.items()]):
raise ValueError(
'At least one metadata group has no samples. Check ' +
'that the mapping file has at least one sample for each value in '
+ 'the passed category.')
if opts.test in TWO_GROUP_TESTS and len(cat_sam_indices) > 2:
option_parser.error(
'The t-test and mann_whitney_u test may ' +
'only be used when there are two sample groups. Choose another ' +
'test or another metadata category.')
# check that assumptions are met for a given test:
if opts.test == 'mann_whitney_u':
sams = reduce(lambda x, y: len(x) + len(y), cat_sam_indices.values())
if sams <= 20:
raise ValueError(
'The number of samples is too small to use the ' +
'Mann-Whitney-U normal approximation. Review the script ' +
'documentation.')
# check that the G-test was not selected if the table appears to be
# relative abundance
if opts.test == 'g_test':
if allclose(bt.sum(axis='sample'), 1.) or (bt.sum(axis='whole') == 1.):
raise ValueError(
'It appears that the biom table you have passed '
'is a relative abundance table where values i,j (obsevation i '
'count in sample j) are fractional and the sum of the columns '
'is 1.0. This will fail to work properly with the G-test. If '
'your data sums to 1 in each column but your data is not '
'relative abundance then the tests will fail anyway because '
'of the reduced number of observations.')
# run actual tests
data_feed = group_significance_row_generator(bt, cat_sam_indices)
test_stats, pvals, means = run_group_significance_test(
data_feed, opts.test, GROUP_TEST_CHOICES, int(opts.permutations))
# calculate corrected pvals
fdr_pvals = array(benjamini_hochberg_step_down(pvals))
bon_pvals = bonferroni_correction(pvals)
# correct for cases where values above 1.0 due to correction
fdr_pvals = where(fdr_pvals > 1.0, 1.0, fdr_pvals)
bon_pvals = where(bon_pvals > 1.0, 1.0, bon_pvals)
# write output results after sorting
lines = group_significance_output_formatter(bt,
test_stats,
pvals,
fdr_pvals,
bon_pvals,
means,
cat_sam_indices,
md_key=opts.metadata_key)
lines = sort_by_pval(lines, ind=2)
o = open(opts.output_fp, 'w')
o.writelines('\n'.join(lines))
o.close()