def test_multiple_covariates_scvi(save_path):
adata = synthetic_iid()
adata.obs["cont1"] = np.random.normal(size=(adata.shape[0], ))
adata.obs["cont2"] = np.random.normal(size=(adata.shape[0], ))
adata.obs["cat1"] = np.random.randint(0, 5, size=(adata.shape[0], ))
adata.obs["cat2"] = np.random.randint(0, 5, size=(adata.shape[0], ))
SCVI.setup_anndata(
adata,
batch_key="batch",
labels_key="labels",
continuous_covariate_keys=["cont1", "cont2"],
categorical_covariate_keys=["cat1", "cat2"],
)
m = SCVI(adata)
m.train(1)
m = SCANVI(adata, unlabeled_category="Unknown")
m.train(1)
TOTALVI.setup_anndata(
adata,
batch_key="batch",
protein_expression_obsm_key="protein_expression",
protein_names_uns_key="protein_names",
continuous_covariate_keys=["cont1", "cont2"],
categorical_covariate_keys=["cat1", "cat2"],
)
m = TOTALVI(adata)
m.train(1)
def test_totalvi_model_library_size(save_path):
adata = synthetic_iid()
TOTALVI.setup_anndata(
adata,
batch_key="batch",
protein_expression_obsm_key="protein_expression",
protein_names_uns_key="protein_names",
)
n_latent = 10
model = TOTALVI(adata, n_latent=n_latent, use_observed_lib_size=False)
assert hasattr(model.module, "library_log_means") and hasattr(
model.module, "library_log_vars")
model.train(1, train_size=0.5)
assert model.is_trained is True
model.get_elbo()
model.get_marginal_ll(n_mc_samples=3)
model.get_latent_library_size()
def test_totalvi_online_update(save_path):
# basic case
n_latent = 5
adata1 = synthetic_iid()
TOTALVI.setup_anndata(
adata1,
batch_key="batch",
protein_expression_obsm_key="protein_expression",
protein_names_uns_key="protein_names",
)
model = TOTALVI(adata1, n_latent=n_latent, use_batch_norm="decoder")
model.train(1, check_val_every_n_epoch=1)
dir_path = os.path.join(save_path, "saved_model/")
model.save(dir_path, overwrite=True)
adata2 = synthetic_iid()
adata2.obs["batch"] = adata2.obs.batch.cat.rename_categories(
["batch_2", "batch_3"])
model2 = TOTALVI.load_query_data(adata2, dir_path)
assert model2.module.background_pro_alpha.requires_grad is True
model2.train(max_epochs=1)
model2.get_latent_representation()
# batch 3 has no proteins
adata2 = synthetic_iid()
adata2.obs["batch"] = adata2.obs.batch.cat.rename_categories(
["batch_2", "batch_3"])
adata2.obsm["protein_expression"][adata2.obs.batch == "batch_3"] = 0
# load from model in memory
model3 = TOTALVI.load_query_data(adata2, model)
model3.module.protein_batch_mask[2]
model3.module.protein_batch_mask[3]
model3.train(max_epochs=1)
model3.get_latent_representation()
def test_totalvi_size_factor():
adata = synthetic_iid()
adata.obs["size_factor"] = np.random.randint(1, 5, size=(adata.shape[0], ))
TOTALVI.setup_anndata(
adata,
batch_key="batch",
protein_expression_obsm_key="protein_expression",
protein_names_uns_key="protein_names",
size_factor_key="size_factor",
)
n_latent = 10
# Test size_factor_key overrides use_observed_lib_size.
model = TOTALVI(adata, n_latent=n_latent, use_observed_lib_size=False)
assert not hasattr(model.module, "library_log_means") and not hasattr(
model.module, "library_log_vars")
assert model.module.use_size_factor_key
model.train(1, train_size=0.5)
model = TOTALVI(adata, n_latent=n_latent, use_observed_lib_size=True)
assert not hasattr(model.module, "library_log_means") and not hasattr(
model.module, "library_log_vars")
assert model.module.use_size_factor_key
model.train(1, train_size=0.5)
def test_totalvi(save_path):
adata = synthetic_iid(run_setup_anndata=False)
TOTALVI.setup_anndata(
adata,
batch_key="batch",
protein_expression_obsm_key="protein_expression",
protein_names_uns_key="protein_names",
)
n_obs = adata.n_obs
n_vars = adata.n_vars
n_proteins = adata.obsm["protein_expression"].shape[1]
n_latent = 10
model = TOTALVI(adata, n_latent=n_latent)
model.train(1, train_size=0.5)
assert model.is_trained is True
z = model.get_latent_representation()
assert z.shape == (n_obs, n_latent)
model.get_elbo()
model.get_marginal_ll(n_mc_samples=3)
model.get_reconstruction_error()
model.get_normalized_expression()
model.get_normalized_expression(transform_batch=["batch_0", "batch_1"])
model.get_latent_library_size()
model.get_protein_foreground_probability()
model.get_protein_foreground_probability(
transform_batch=["batch_0", "batch_1"])
post_pred = model.posterior_predictive_sample(n_samples=2)
assert post_pred.shape == (n_obs, n_vars + n_proteins, 2)
post_pred = model.posterior_predictive_sample(n_samples=1)
assert post_pred.shape == (n_obs, n_vars + n_proteins)
feature_correlation_matrix1 = model.get_feature_correlation_matrix(
correlation_type="spearman")
feature_correlation_matrix1 = model.get_feature_correlation_matrix(
correlation_type="spearman", transform_batch=["batch_0", "batch_1"])
feature_correlation_matrix2 = model.get_feature_correlation_matrix(
correlation_type="pearson")
assert feature_correlation_matrix1.shape == (
n_vars + n_proteins,
n_vars + n_proteins,
)
assert feature_correlation_matrix2.shape == (
n_vars + n_proteins,
n_vars + n_proteins,
)
model.get_elbo(indices=model.validation_indices)
model.get_marginal_ll(indices=model.validation_indices, n_mc_samples=3)
model.get_reconstruction_error(indices=model.validation_indices)
adata2 = synthetic_iid(run_setup_anndata=False)
TOTALVI.setup_anndata(
adata2,
batch_key="batch",
protein_expression_obsm_key="protein_expression",
protein_names_uns_key="protein_names",
)
norm_exp = model.get_normalized_expression(adata2, indices=[1, 2, 3])
assert norm_exp[0].shape == (3, adata2.n_vars)
assert norm_exp[1].shape == (3, adata2.obsm["protein_expression"].shape[1])
latent_lib_size = model.get_latent_library_size(adata2, indices=[1, 2, 3])
assert latent_lib_size.shape == (3, 1)
pro_foreground_prob = model.get_protein_foreground_probability(
adata2, indices=[1, 2, 3], protein_list=["1", "2"])
assert pro_foreground_prob.shape == (3, 2)
model.posterior_predictive_sample(adata2)
model.get_feature_correlation_matrix(adata2)
# test transfer_anndata_setup + view
adata2 = synthetic_iid(run_setup_anndata=False)
transfer_anndata_setup(adata, adata2)
model.get_elbo(adata2[:10])
# test automatic transfer_anndata_setup
adata = synthetic_iid()
model = TOTALVI(adata)
adata2 = synthetic_iid(run_setup_anndata=False)
model.get_elbo(adata2)
# test that we catch incorrect mappings
adata = synthetic_iid()
adata2 = synthetic_iid(run_setup_anndata=False)
transfer_anndata_setup(adata, adata2)
adata2.uns["_scvi"]["categorical_mappings"]["_scvi_labels"][
"mapping"] = np.array(["label_1", "label_0", "label_8"])
with pytest.raises(ValueError):
model.get_elbo(adata2)
# test that same mapping different order is okay
adata = synthetic_iid()
adata2 = synthetic_iid(run_setup_anndata=False)
transfer_anndata_setup(adata, adata2)
adata2.uns["_scvi"]["categorical_mappings"]["_scvi_labels"][
"mapping"] = np.array(["label_1", "label_0", "label_2"])
model.get_elbo(adata2) # should automatically transfer setup
# test that we catch missing proteins
adata2 = synthetic_iid(run_setup_anndata=False)
del adata2.obsm["protein_expression"]
with pytest.raises(KeyError):
model.get_elbo(adata2)
model.differential_expression(groupby="labels", group1="label_1")
model.differential_expression(groupby="labels",
group1="label_1",
group2="label_2")
model.differential_expression(idx1=[0, 1, 2], idx2=[3, 4, 5])
model.differential_expression(idx1=[0, 1, 2])
model.differential_expression(groupby="labels")
# test with missing proteins
adata = scvi.data.pbmcs_10x_cite_seq(save_path=save_path,
protein_join="outer")
model = TOTALVI(adata)
assert model.module.protein_batch_mask is not None
model.train(1, train_size=0.5)
model = TOTALVI(adata, override_missing_proteins=True)
assert model.module.protein_batch_mask is None
model.train(1, train_size=0.5)