def test_dense_subsample_genes(self):
data = [
np.random.randint(1, 5, size=(50, 26)),
np.random.randint(1, 5, size=(50, 26)),
np.random.randint(1, 5, size=(50, 26)),
]
# With default
dataset = GeneExpressionDataset()
dataset.populate_from_per_batch_list(data)
n_genes = dataset.nb_genes
n_top = n_genes // 2
dataset.subsample_genes(new_n_genes=n_top)
assert dataset.nb_genes < n_genes
# For some reason the new number of genes can be slightly different than n_top
# With Seurat
dataset = GeneExpressionDataset()
dataset.populate_from_per_batch_list(data)
dataset.subsample_genes(new_n_genes=n_top, mode="seurat")
assert dataset.nb_genes < n_genes
def test_batch_correction(self):
data = [
np.random.randint(1, 5, size=(50, 25)),
np.random.randint(1, 5, size=(50, 25)),
np.random.randint(1, 5, size=(50, 25)),
]
dataset = GeneExpressionDataset()
dataset.populate_from_per_batch_list(data)
n_genes = dataset.nb_genes
n_top = n_genes // 2
dataset._highly_variable_genes(n_bins=3, flavor="seurat_v2")
df = dataset._highly_variable_genes(
n_bins=3, n_top_genes=n_top, flavor="seurat_v2"
)
assert df["highly_variable"].sum() >= n_top
dataset.subsample_genes(new_n_genes=n_top)
new_genes = dataset.nb_genes
assert n_genes > new_genes, "subsample_genes did not filter out genes"
pass
def test_dense_subsample_genes(self):
data = [
np.random.randint(1, 5, size=(50, 26)),
np.random.randint(1, 5, size=(50, 26)),
np.random.randint(1, 5, size=(50, 26)),
]
# With default
dataset = GeneExpressionDataset()
dataset.populate_from_per_batch_list(data)
n_genes = dataset.nb_genes
n_top = n_genes // 2
dataset.subsample_genes(new_n_genes=n_top, mode="cell_ranger")
assert dataset.nb_genes == n_top
# With Seurat v2
dataset = GeneExpressionDataset()
dataset.populate_from_per_batch_list(data)
dataset.subsample_genes(new_n_genes=n_top, mode="seurat_v2")
assert dataset.nb_genes == n_top
# With Seurat v3
dataset = GeneExpressionDataset()
dataset.populate_from_per_batch_list(data)
dataset.subsample_genes(new_n_genes=n_top, mode="seurat_v3")
assert dataset.nb_genes == n_top
def test_subsample_genes(self):
data = np.ones((25, 100)) * 100
variable_data = data
variable_data[0, :] = 2
variable_data *= np.arange(0, 100)
gene_names = np.array(["gene_%d" % i for i in range(100)])
dataset = GeneExpressionDataset()
dataset.populate_from_data(data, gene_names=gene_names)
dataset.subsample_genes(new_ratio_genes=0.4, mode="variance")
self.assertTupleEqual(dataset.gene_names.shape, (40, ))
dataset.subsample_genes(new_n_genes=25, mode="variance")
self.assertTupleEqual(dataset.gene_names.shape, (25, ))
# The most variable genes should be in first position
self.assertEqual(dataset.gene_names[0], "GENE_99")
dataset.subsample_genes(subset_genes=[1, 6, 7])
self.assertEqual(dataset.gene_names[0], "GENE_98")
#initial number of sampled cells
sampling_size = 500
while sampling_size < n_retained_cells:
cells_sizes.append(sampling_size)
sampling_size = int(sampling_size*np.sqrt(2))
# cells_sizes = np.logspace(np.log2(500), np.log2(n_retained_cells), num=9, base=2).astype(int)
print('Number of sampled cells for ', ds, cells_sizes)
cells_dataset = GeneExpressionDataset()
X_ = adata.layers['0']
cells_dataset.populate_from_data(X_, gene_names=adata.var.index.values)
#we subsambple to 1000 genes for speed and to prevent overfitting
cells_dataset.filter_genes_by_count(per_batch=True)
cells_dataset.subsample_genes(1000)
sel_genes = cells_dataset.gene_names
n_validation = adata.shape[0] - n_retained_cells
print(ds, ' n_validation:', n_validation)
validation_cells = np.random.choice(adata.obs.index, size=n_validation, replace=False)
learning_cells = adata.obs.index.difference(validation_cells)
val_adata = adata[validation_cells]
lea_adata = adata[learning_cells]
ne_cells = X_.sum(axis=1) > 0
to_keep = np.where(ne_cells)[0]
log_counts = np.log(X_[to_keep].sum(axis=1))
