def test_gtex_reindex(expression_data):
"""Test that all of the transforms reindex to GTEx properly."""
identifier = 'symbol'
norm = normalize.Normalizer(identifier=identifier)
tpm = norm.tpm_from_counts(expression_data.counts)
clr = norm.clr_from_tpm(tpm, imputer=normalize.impute)
assert (tpm.columns == clr.columns).all()
assert (tpm.columns == norm.gene_lengths.index).all()
def test_clr_functions(expression_data):
"""Test the TPM -> CLR and CLR -> TPM transforms for some expression data."""
identifier = 'symbol'
norm = normalize.Normalizer(identifier=identifier)
tpm = normalize.impute(expression_data.tpm)
clr = norm.clr_from_tpm(tpm, gene_list=tpm.columns)
tpm_from_clr = norm.tpm_from_clr(clr, gene_list=clr.columns)
assert np.allclose(tpm, tpm_from_clr)
def test_normalizer_tpm_from_rpkm_allgenes(expression_data):
"""Test the RPKM -> TPM conversion for some expression data.
Include all genes."""
identifier = 'symbol'
norm = normalize.Normalizer(identifier=identifier)
rpkm = expression_data.rpkm
tpm = expression_data.tpm
tpm_calc = norm.tpm_from_rpkm(rpkm)
assert (tpm.index == tpm_calc.index).all()
assert np.allclose(tpm.values, tpm_calc[expression_data.tpm.columns].values)
def test_zscore_from_clr(expression_data):
"""Test the z-score transformation on CLR data."""
identifier = 'symbol'
norm = normalize.Normalizer(identifier=identifier)
tpm = normalize.impute(expression_data.tpm)
clr = norm.clr_from_tpm(tpm, gene_list=tpm.columns)
tissues = pd.Series('Liver', index=clr.index)
zscore = norm.z_score_from_clr(clr, tissues)
assert zscore.shape == clr.shape
def test_normalizer_tpm_from_counts(expression_data):
"""Test the counts -> TPM conversion for some expression data."""
identifier = 'symbol'
norm = normalize.Normalizer(identifier=identifier)
counts = expression_data.counts
tpm = expression_data.tpm
tpm_calc = norm.tpm_from_counts(counts)
assert (tpm.columns == tpm_calc.columns).all()
assert (tpm.index == tpm_calc.index).all()
assert np.allclose(tpm.values, tpm_calc.values)
def test_normalizer_tpm_from_subset(expression_data):
"""Test the TPM -> TPM subset conversion for some expression data."""
identifier = 'symbol'
norm = normalize.Normalizer(identifier=identifier)
tpm = expression_data.tpm
tpm_fullset_calc = norm.tpm_from_subset(tpm)
assert np.allclose(tpm.values, tpm_fullset_calc.values)
tpm_subset_calc = norm.tpm_from_subset(tpm, tpm.columns[:100])
tpm_subset_norm = tpm_subset_calc.sum(axis=1).values
assert np.allclose(tpm_subset_norm - 1e6, np.zeros_like(tpm_subset_norm))
def test_ordinalize(expression_data):
"""Test the ordinalize transformation on CLR data."""
identifier = 'symbol'
norm = normalize.Normalizer(identifier=identifier)
tpm = normalize.impute(expression_data.tpm)
clr = norm.clr_from_tpm(tpm, gene_list=tpm.columns)
cutoffs = [0.37]
min_value = 5
ords = norm.ordinalize(clr, cutoffs, min_value=min_value)
assert ((clr <= cutoffs[0]) == (ords == min_value)).all().all()
assert ((clr > cutoffs[0]) == (ords == 1+min_value)).all().all()
def test_alr_functions_dirimpute(expression_data):
"""Test the TPM -> ALR transform for some expression data.
Directly impute in the alr calculation."""
identifier = 'symbol'
norm = normalize.Normalizer(identifier=identifier)
all_genes = list(expression_data.tpm.columns)
reference_genes = all_genes[:1]
genes_to_keep = all_genes[1:]
alr = norm.alr_from_tpm(expression_data.tpm, reference_genes,
gene_list=all_genes, imputer=normalize.impute)
alr_genes = list(alr.columns)
assert (genes_to_keep == alr_genes)
def expression_data():
"""Create some example data."""
num_samples = 100
num_genes = 1000
max_read_count = 1234
counts = pd.DataFrame(np.round(max_read_count*np.random.rand(num_samples, num_genes)))
# create a Normalizer object and get gene lengths
norm = normalize.Normalizer(identifier='symbol')
gene_lengths = norm.gene_lengths[:num_genes]
counts.columns = gene_lengths.index
# TPM
tpk = counts.divide(gene_lengths/1e3, axis='columns')
tpm = tpk.divide(tpk.sum(axis=1)/1e6, axis='index')
# RPKM
cpm = counts.divide(counts.sum(axis=1)/1e6, axis='index')
rpkm = cpm.divide(gene_lengths/1e3, axis='columns')
return ExpressionData(counts, tpm, rpkm)
