scvi.model.MULTIVI#

class scvi.model.MULTIVI(adata, n_genes, n_regions, modality_weights='equal', modality_penalty='Jeffreys', n_hidden=None, n_latent=None, n_layers_encoder=2, n_layers_decoder=2, dropout_rate=0.1, region_factors=True, gene_likelihood='zinb', dispersion='gene', use_batch_norm='none', use_layer_norm='both', latent_distribution='normal', deeply_inject_covariates=False, encode_covariates=False, fully_paired=False, protein_dispersion='protein', **model_kwargs)[source]#

Integration of multi-modal and single-modality data [Ashuach et al., 2023].

MultiVI is used to integrate multiomic datasets with single-modality (expression or accessibility) datasets.

Parameters:
  • adata (AnnData) – AnnData object that has been registered via setup_anndata().

  • n_genes (int) – The number of gene expression features (genes).

  • n_regions (int) – The number of accessibility features (genomic regions).

  • modality_weights (Literal['equal', 'cell', 'universal'] (default: 'equal')) – Weighting scheme across modalities. One of the following: * "equal": Equal weight in each modality * "universal": Learn weights across modalities w_m. * "cell": Learn weights across modalities and cells. w_{m,c}

  • modality_penalty (Literal['Jeffreys', 'MMD', 'None'] (default: 'Jeffreys')) – Training Penalty across modalities. One of the following: * "Jeffreys": Jeffreys penalty to align modalities * "MMD": MMD penalty to align modalities * "None": No penalty

  • n_hidden (int | None (default: None)) – Number of nodes per hidden layer. If None, defaults to square root of number of regions.

  • n_latent (int | None (default: None)) – Dimensionality of the latent space. If None, defaults to square root of n_hidden.

  • n_layers_encoder (int (default: 2)) – Number of hidden layers used for encoder NNs.

  • n_layers_decoder (int (default: 2)) – Number of hidden layers used for decoder NNs.

  • dropout_rate (float (default: 0.1)) – Dropout rate for neural networks.

  • model_depth – Model sequencing depth / library size.

  • region_factors (bool (default: True)) – Include region-specific factors in the model.

  • gene_dispersion – One of the following * 'gene' - genes_dispersion parameter of NB is constant per gene across cells * 'gene-batch' - genes_dispersion can differ between different batches * 'gene-label' - genes_dispersion can differ between different labels

  • protein_dispersion (Literal['protein', 'protein-batch', 'protein-label'] (default: 'protein')) – One of the following * 'protein' - protein_dispersion parameter is constant per protein across cells * 'protein-batch' - protein_dispersion can differ between different batches NOT TESTED * 'protein-label' - protein_dispersion can differ between different labels NOT TESTED

  • latent_distribution (Literal['normal', 'ln'] (default: 'normal')) – One of * 'normal' - Normal distribution * 'ln' - Logistic normal distribution (Normal(0, I) transformed by softmax)

  • deeply_inject_covariates (bool (default: False)) – Whether to deeply inject covariates into all layers of the decoder. If False, covariates will only be included in the input layer.

  • fully_paired (bool (default: False)) – allows the simplification of the model if the data is fully paired. Currently ignored.

  • **model_kwargs – Keyword args for MULTIVAE

Examples

>>> adata_rna = anndata.read_h5ad(path_to_rna_anndata)
>>> adata_atac = scvi.data.read_10x_atac(path_to_atac_anndata)
>>> adata_multi = scvi.data.read_10x_multiome(path_to_multiomic_anndata)
>>> adata_mvi = scvi.data.organize_multiome_anndatas(adata_multi, adata_rna, adata_atac)
>>> scvi.model.MULTIVI.setup_anndata(adata_mvi, batch_key="modality")
>>> vae = scvi.model.MULTIVI(adata_mvi)
>>> vae.train()

Notes

  • The model assumes that the features are organized so that all expression features are

    consecutive, followed by all accessibility features. For example, if the data has 100 genes and 250 genomic regions, the model assumes that the first 100 features are genes, and the next 250 are the regions.

  • The main batch annotation, specified in setup_anndata, should correspond to

    the modality each cell originated from. This allows the model to focus mixing efforts, using an adversarial component, on mixing the modalities. Other covariates can be specified using the categorical_covariate_keys argument.

Attributes table#

adata

Data attached to model instance.

adata_manager

Manager instance associated with self.adata.

device

The current device that the module's params are on.

history

Returns computed metrics during training.

is_trained

Whether the model has been trained.

summary_string

Summary string of the model.

test_indices

Observations that are in test set.

train_indices

Observations that are in train set.

validation_indices

Observations that are in validation set.

Methods table#

convert_legacy_save(dir_path, output_dir_path)

Converts a legacy saved model (<v0.15.0) to the updated save format.

deregister_manager([adata])

Deregisters the AnnDataManager instance associated with adata.

differential_accessibility([adata, groupby, ...])

A unified method for differential accessibility analysis.

differential_expression([adata, groupby, ...])

A unified method for differential expression analysis.

get_accessibility_estimates([adata, ...])

Impute the full accessibility matrix.

get_anndata_manager(adata[, required])

Retrieves the AnnDataManager for a given AnnData object specific to this model instance.

get_elbo([adata, indices, batch_size])

Return the ELBO for the data.

get_from_registry(adata, registry_key)

Returns the object in AnnData associated with the key in the data registry.

get_latent_representation([adata, modality, ...])

Return the latent representation for each cell.

get_library_size_factors([adata, indices, ...])

Return library size factors.

get_marginal_ll([adata, indices, ...])

Return the marginal LL for the data.

get_normalized_expression([adata, indices, ...])

Returns the normalized (decoded) gene expression.

get_protein_foreground_probability([adata, ...])

Returns the foreground probability for proteins.

get_reconstruction_error([adata, indices, ...])

Return the reconstruction error for the data.

get_region_factors()

Return region-specific factors.

load(dir_path[, adata, accelerator, device, ...])

Instantiate a model from the saved output.

load_query_data(adata, reference_model[, ...])

Online update of a reference model with scArches algorithm [Lotfollahi et al., 2021].

load_registry(dir_path[, prefix])

Return the full registry saved with the model.

prepare_query_anndata(adata, reference_model)

Prepare data for query integration.

register_manager(adata_manager)

Registers an AnnDataManager instance with this model class.

save(dir_path[, prefix, overwrite, ...])

Save the state of the model.

setup_anndata(adata[, layer, batch_key, ...])

Sets up the AnnData object for this model.

to_device(device)

Move model to device.

train([max_epochs, lr, accelerator, ...])

Trains the model using amortized variational inference.

view_anndata_setup([adata, ...])

Print summary of the setup for the initial AnnData or a given AnnData object.

view_setup_args(dir_path[, prefix])

Print args used to setup a saved model.

Attributes#

MULTIVI.adata[source]#

Data attached to model instance.

MULTIVI.adata_manager[source]#

Manager instance associated with self.adata.

MULTIVI.device[source]#

The current device that the module’s params are on.

MULTIVI.history[source]#

Returns computed metrics during training.

MULTIVI.is_trained[source]#

Whether the model has been trained.

MULTIVI.summary_string[source]#

Summary string of the model.

MULTIVI.test_indices[source]#

Observations that are in test set.

MULTIVI.train_indices[source]#

Observations that are in train set.

MULTIVI.validation_indices[source]#

Observations that are in validation set.

Methods#

classmethod MULTIVI.convert_legacy_save(dir_path, output_dir_path, overwrite=False, prefix=None, **save_kwargs)[source]#

Converts a legacy saved model (<v0.15.0) to the updated save format.

Parameters:
  • dir_path (str) – Path to directory where legacy model is saved.

  • output_dir_path (str) – Path to save converted save files.

  • overwrite (bool (default: False)) – Overwrite existing data or not. If False and directory already exists at output_dir_path, error will be raised.

  • prefix (str | None (default: None)) – Prefix of saved file names.

  • **save_kwargs – Keyword arguments passed into save().

Return type:

None

MULTIVI.deregister_manager(adata=None)[source]#

Deregisters the AnnDataManager instance associated with adata.

If adata is None, deregisters all AnnDataManager instances in both the class and instance-specific manager stores, except for the one associated with this model instance.

MULTIVI.differential_accessibility(adata=None, groupby=None, group1=None, group2=None, idx1=None, idx2=None, mode='change', delta=0.05, batch_size=None, all_stats=True, batch_correction=False, batchid1=None, batchid2=None, fdr_target=0.05, silent=False, two_sided=True, **kwargs)[source]#

A unified method for differential accessibility analysis.

Implements 'vanilla' DE [Lopez et al., 2018] and 'change' mode DE [Boyeau et al., 2019].

Parameters:
  • adata (AnnData | None (default: None)) – AnnData object with equivalent structure to initial AnnData. If None, defaults to the AnnData object used to initialize the model.

  • groupby (str | None (default: None)) – The key of the observations grouping to consider.

  • group1 (Iterable[str] | None (default: None)) – Subset of groups, e.g. [‘g1’, ‘g2’, ‘g3’], to which comparison shall be restricted, or all groups in groupby (default).

  • group2 (str | None (default: None)) – If None, compare each group in group1 to the union of the rest of the groups in groupby. If a group identifier, compare with respect to this group.

  • idx1 (Sequence[int] | Sequence[bool] | None (default: None)) – idx1 and idx2 can be used as an alternative to the AnnData keys. Custom identifier for group1 that can be of three sorts: (1) a boolean mask, (2) indices, or (3) a string. If it is a string, then it will query indices that verifies conditions on adata.obs, as described in pandas.DataFrame.query() If idx1 is not None, this option overrides group1 and group2.

  • idx2 (Sequence[int] | Sequence[bool] | None (default: None)) – Custom identifier for group2 that has the same properties as idx1. By default, includes all cells not specified in idx1.

  • mode (Literal['vanilla', 'change'] (default: 'change')) – Method for differential expression. See user guide for full explanation.

  • delta (float (default: 0.05)) – specific case of region inducing differential expression. In this case, we suppose that \(R \setminus [-\delta, \delta]\) does not induce differential expression (change model default case).

  • batch_size (int | None (default: None)) – Minibatch size for data loading into model. Defaults to scvi.settings.batch_size.

  • all_stats (bool (default: True)) – Concatenate count statistics (e.g., mean expression group 1) to DE results.

  • batch_correction (bool (default: False)) – Whether to correct for batch effects in DE inference.

  • batchid1 (Iterable[str] | None (default: None)) – Subset of categories from batch_key registered in setup_anndata, e.g. [‘batch1’, ‘batch2’, ‘batch3’], for group1. Only used if batch_correction is True, and by default all categories are used.

  • batchid2 (Iterable[str] | None (default: None)) – Same as batchid1 for group2. batchid2 must either have null intersection with batchid1, or be exactly equal to batchid1. When the two sets are exactly equal, cells are compared by decoding on the same batch. When sets have null intersection, cells from group1 and group2 are decoded on each group in group1 and group2, respectively.

  • fdr_target (float (default: 0.05)) – Tag features as DE based on posterior expected false discovery rate.

  • silent (bool (default: False)) – If True, disables the progress bar. Default: False.

  • two_sided (bool (default: True)) – Whether to perform a two-sided test, or a one-sided test.

  • **kwargs – Keyword args for scvi.model.base.DifferentialComputation.get_bayes_factors()

Return type:

DataFrame

Returns:

Differential accessibility DataFrame with the following columns: prob_da

the probability of the region being differentially accessible

is_da_fdr

whether the region passes a multiple hypothesis correction procedure with the target_fdr threshold

bayes_factor

Bayes Factor indicating the level of significance of the analysis

effect_size

the effect size, computed as (accessibility in population 2) - (accessibility in population 1)

emp_effect

the empirical effect, based on observed detection rates instead of the estimated accessibility scores from the PeakVI model

est_prob1

the estimated probability of accessibility in population 1

est_prob2

the estimated probability of accessibility in population 2

emp_prob1

the empirical (observed) probability of accessibility in population 1

emp_prob2

the empirical (observed) probability of accessibility in population 2

MULTIVI.differential_expression(adata=None, groupby=None, group1=None, group2=None, idx1=None, idx2=None, mode='change', delta=0.25, batch_size=None, all_stats=True, batch_correction=False, batchid1=None, batchid2=None, fdr_target=0.05, silent=False, **kwargs)[source]#

A unified method for differential expression analysis.

Implements “vanilla” DE [Lopez et al., 2018] and “change” mode DE [Boyeau et al., 2019].

Parameters:
  • adata (AnnData | None (default: None)) – AnnData object with equivalent structure to initial AnnData. If None, defaults to the AnnData object used to initialize the model.

  • groupby (str | None (default: None)) – The key of the observations grouping to consider.

  • group1 (Iterable[str] | None (default: None)) – Subset of groups, e.g. [‘g1’, ‘g2’, ‘g3’], to which comparison shall be restricted, or all groups in groupby (default).

  • group2 (str | None (default: None)) – If None, compare each group in group1 to the union of the rest of the groups in groupby. If a group identifier, compare with respect to this group.

  • idx1 (Sequence[int] | Sequence[bool] | None (default: None)) – idx1 and idx2 can be used as an alternative to the AnnData keys. Custom identifier for group1 that can be of three sorts: (1) a boolean mask, (2) indices, or (3) a string. If it is a string, then it will query indices that verifies conditions on adata.obs, as described in pandas.DataFrame.query() If idx1 is not None, this option overrides group1 and group2.

  • idx2 (Sequence[int] | Sequence[bool] | None (default: None)) – Custom identifier for group2 that has the same properties as idx1. By default, includes all cells not specified in idx1.

  • mode (Literal['vanilla', 'change'] (default: 'change')) – Method for differential expression. See user guide for full explanation.

  • delta (float (default: 0.25)) – specific case of region inducing differential expression. In this case, we suppose that \(R \setminus [-\delta, \delta]\) does not induce differential expression (change model default case).

  • batch_size (int | None (default: None)) – Minibatch size for data loading into model. Defaults to scvi.settings.batch_size.

  • all_stats (bool (default: True)) – Concatenate count statistics (e.g., mean expression group 1) to DE results.

  • batch_correction (bool (default: False)) – Whether to correct for batch effects in DE inference.

  • batchid1 (Iterable[str] | None (default: None)) – Subset of categories from batch_key registered in setup_anndata, e.g. [‘batch1’, ‘batch2’, ‘batch3’], for group1. Only used if batch_correction is True, and by default all categories are used.

  • batchid2 (Iterable[str] | None (default: None)) – Same as batchid1 for group2. batchid2 must either have null intersection with batchid1, or be exactly equal to batchid1. When the two sets are exactly equal, cells are compared by decoding on the same batch. When sets have null intersection, cells from group1 and group2 are decoded on each group in group1 and group2, respectively.

  • fdr_target (float (default: 0.05)) – Tag features as DE based on posterior expected false discovery rate.

  • silent (bool (default: False)) – If True, disables the progress bar. Default: False.

  • **kwargs – Keyword args for scvi.model.base.DifferentialComputation.get_bayes_factors()

Return type:

DataFrame

Returns:

Differential expression DataFrame.

MULTIVI.get_accessibility_estimates(adata=None, indices=None, n_samples_overall=None, region_list=None, transform_batch=None, use_z_mean=True, threshold=None, normalize_cells=False, normalize_regions=False, batch_size=128, return_numpy=False)[source]#

Impute the full accessibility matrix.

Returns a matrix of accessibility probabilities for each cell and genomic region in the input (for return matrix A, A[i,j] is the probability that region j is accessible in cell i).

Parameters:
  • adata (AnnData | None (default: None)) – AnnData object that has been registered with scvi. If None, defaults to the AnnData object used to initialize the model.

  • indices (Sequence[int] (default: None)) – Indices of cells in adata to use. If None, all cells are used.

  • n_samples_overall (int | None (default: None)) – Number of samples to return in total

  • region_list (Sequence[str] | None (default: None)) – Regions to use. if None, all regions are used.

  • transform_batch (str | int | None (default: None)) –

    Batch to condition on. If transform_batch is:

    • None, then real observed batch is used

    • int, then batch transform_batch is used

  • use_z_mean (bool (default: True)) – If True (default), use the distribution mean. Otherwise, sample from the distribution.

  • threshold (float | None (default: None)) – If provided, values below the threshold are replaced with 0 and a sparse matrix is returned instead. This is recommended for very large matrices. Must be between 0 and 1.

  • normalize_cells (bool (default: False)) – Whether to reintroduce library size factors to scale the normalized probabilities. This makes the estimates closer to the input, but removes the library size correction. False by default.

  • normalize_regions (bool (default: False)) – Whether to reintroduce region factors to scale the normalized probabilities. This makes the estimates closer to the input, but removes the region-level bias correction. False by default.

  • batch_size (int (default: 128)) – Minibatch size for data loading into model

Return type:

ndarray | csr_matrix | DataFrame

MULTIVI.get_anndata_manager(adata, required=False)[source]#

Retrieves the AnnDataManager for a given AnnData object specific to this model instance.

Requires self.id has been set. Checks for an AnnDataManager specific to this model instance.

Parameters:
  • adata (Union[AnnData, MuData]) – AnnData object to find manager instance for.

  • required (bool (default: False)) – If True, errors on missing manager. Otherwise, returns None when manager is missing.

Return type:

AnnDataManager | None

MULTIVI.get_elbo(adata=None, indices=None, batch_size=None)[source]#

Return the ELBO for the data.

The ELBO is a lower bound on the log likelihood of the data used for optimization of VAEs. Note, this is not the negative ELBO, higher is better.

Parameters:
  • adata (Optional[AnnData] (default: None)) – AnnData object with equivalent structure to initial AnnData. If None, defaults to the AnnData object used to initialize the model.

  • indices (Optional[Sequence[int]] (default: None)) – Indices of cells in adata to use. If None, all cells are used.

  • batch_size (Optional[int] (default: None)) – Minibatch size for data loading into model. Defaults to scvi.settings.batch_size.

Return type:

float

MULTIVI.get_from_registry(adata, registry_key)[source]#

Returns the object in AnnData associated with the key in the data registry.

AnnData object should be registered with the model prior to calling this function via the self._validate_anndata method.

Parameters:
  • registry_key (str) – key of object to get from data registry.

  • adata (Union[AnnData, MuData]) – AnnData to pull data from.

Return type:

ndarray

Returns:

The requested data as a NumPy array.

MULTIVI.get_latent_representation(adata=None, modality='joint', indices=None, give_mean=True, batch_size=None)[source]#

Return the latent representation for each cell.

Parameters:
  • adata (AnnData | None (default: None)) – AnnData object with equivalent structure to initial AnnData. If None, defaults to the AnnData object used to initialize the model.

  • modality (Literal['joint', 'expression', 'accessibility'] (default: 'joint')) – Return modality specific or joint latent representation.

  • indices (Sequence[int] | None (default: None)) – Indices of cells in adata to use. If None, all cells are used.

  • give_mean (bool (default: True)) – Give mean of distribution or sample from it.

  • batch_size (int | None (default: None)) – Minibatch size for data loading into model. Defaults to scvi.settings.batch_size.

Return type:

ndarray

Returns:

-latent_representation (ndarray) Low-dimensional representation for each cell

MULTIVI.get_library_size_factors(adata=None, indices=None, batch_size=128)[source]#

Return library size factors.

Parameters:
  • adata (AnnData | None (default: None)) – AnnData object with equivalent structure to initial AnnData. If None, defaults to the AnnData object used to initialize the model.

  • indices (Sequence[int] (default: None)) – Indices of cells in adata to use. If None, all cells are used.

  • batch_size (int (default: 128)) – Minibatch size for data loading into model. Defaults to scvi.settings.batch_size.

Return type:

dict[str, ndarray]

Returns:

Library size factor for expression and accessibility

MULTIVI.get_marginal_ll(adata=None, indices=None, n_mc_samples=1000, batch_size=None, return_mean=True, **kwargs)[source]#

Return the marginal LL for the data.

The computation here is a biased estimator of the marginal log likelihood of the data. Note, this is not the negative log likelihood, higher is better.

Parameters:
  • adata (Optional[AnnData] (default: None)) – AnnData object with equivalent structure to initial AnnData. If None, defaults to the AnnData object used to initialize the model.

  • indices (Optional[Sequence[int]] (default: None)) – Indices of cells in adata to use. If None, all cells are used.

  • n_mc_samples (int (default: 1000)) – Number of Monte Carlo samples to use for marginal LL estimation.

  • batch_size (Optional[int] (default: None)) – Minibatch size for data loading into model. Defaults to scvi.settings.batch_size.

  • return_mean (Optional[bool] (default: True)) – If False, return the marginal log likelihood for each observation. Otherwise, return the mmean arginal log likelihood.

Return type:

Union[Tensor, float]

MULTIVI.get_normalized_expression(adata=None, indices=None, n_samples_overall=None, transform_batch=None, gene_list=None, use_z_mean=True, n_samples=1, batch_size=None, return_mean=True, return_numpy=False)[source]#

Returns the normalized (decoded) gene expression.

This is denoted as \(\rho_n\) in the scVI paper.

Parameters:
  • adata (AnnData | None (default: None)) – AnnData object with equivalent structure to initial AnnData. If None, defaults to the AnnData object used to initialize the model.

  • indices (Sequence[int] | None (default: None)) – Indices of cells in adata to use. If None, all cells are used.

  • transform_batch (Sequence[Union[int, float, str]] | None (default: None)) –

    Batch to condition on. If transform_batch is:

    • None, then real observed batch is used.

    • int, then batch transform_batch is used.

  • gene_list (Sequence[str] | None (default: None)) – Return frequencies of expression for a subset of genes. This can save memory when working with large datasets and few genes are of interest.

  • library_size – Scale the expression frequencies to a common library size. This allows gene expression levels to be interpreted on a common scale of relevant magnitude. If set to “latent”, use the latent library size.

  • use_z_mean (bool (default: True)) – If True, use the mean of the latent distribution, otherwise sample from it

  • n_samples (int (default: 1)) – Number of posterior samples to use for estimation.

  • batch_size (int | None (default: None)) – Minibatch size for data loading into model. Defaults to scvi.settings.batch_size.

  • return_mean (bool (default: True)) – Whether to return the mean of the samples.

Return type:

ndarray | DataFrame

Returns:

If n_samples > 1 and return_mean is False, then the shape is (samples, cells, genes). Otherwise, shape is (cells, genes). In this case, return type is DataFrame unless return_numpy is True.

MULTIVI.get_protein_foreground_probability(adata=None, indices=None, transform_batch=None, protein_list=None, n_samples=1, batch_size=None, use_z_mean=True, return_mean=True, return_numpy=None)[source]#

Returns the foreground probability for proteins.

This is denoted as \((1 - \pi_{nt})\) in the totalVI paper.

Parameters:
  • adata (AnnData | None (default: None)) – AnnData object with equivalent structure to initial AnnData. If None, defaults to the AnnData object used to initialize the model.

  • indices (Sequence[int] | None (default: None)) – Indices of cells in adata to use. If None, all cells are used.

  • transform_batch (Sequence[Union[int, float, str]] | None (default: None)) –

    Batch to condition on. If transform_batch is:

    • None - real observed batch is used

    • int - batch transform_batch is used

    • List[int] - average over batches in list

  • protein_list (Sequence[str] | None (default: None)) – Return protein expression for a subset of genes. This can save memory when working with large datasets and few genes are of interest.

  • n_samples (int (default: 1)) – Number of posterior samples to use for estimation.

  • batch_size (int | None (default: None)) – Minibatch size for data loading into model. Defaults to scvi.settings.batch_size.

  • return_mean (bool (default: True)) – Whether to return the mean of the samples.

  • return_numpy (bool | None (default: None)) – Return a ndarray instead of a DataFrame. DataFrame includes gene names as columns. If either n_samples=1 or return_mean=True, defaults to False. Otherwise, it defaults to True.

Returns:

  • foreground_probability - probability foreground for each protein

If n_samples > 1 and return_mean is False, then the shape is (samples, cells, genes). Otherwise, shape is (cells, genes). In this case, return type is DataFrame unless return_numpy is True.

MULTIVI.get_reconstruction_error(adata=None, indices=None, batch_size=None)[source]#

Return the reconstruction error for the data.

This is typically written as \(p(x \mid z)\), the likelihood term given one posterior sample. Note, this is not the negative likelihood, higher is better.

Parameters:
  • adata (Optional[AnnData] (default: None)) – AnnData object with equivalent structure to initial AnnData. If None, defaults to the AnnData object used to initialize the model.

  • indices (Optional[Sequence[int]] (default: None)) – Indices of cells in adata to use. If None, all cells are used.

  • batch_size (Optional[int] (default: None)) – Minibatch size for data loading into model. Defaults to scvi.settings.batch_size.

Return type:

float

MULTIVI.get_region_factors()[source]#

Return region-specific factors.

Return type:

ndarray

classmethod MULTIVI.load(dir_path, adata=None, accelerator='auto', device='auto', prefix=None, backup_url=None)[source]#

Instantiate a model from the saved output.

Parameters:
  • dir_path (str) – Path to saved outputs.

  • adata (Union[AnnData, MuData, None] (default: None)) – AnnData organized in the same way as data used to train model. It is not necessary to run setup_anndata, as AnnData is validated against the saved scvi setup dictionary. If None, will check for and load anndata saved with the model.

  • accelerator (str (default: 'auto')) – Supports passing different accelerator types (“cpu”, “gpu”, “tpu”, “ipu”, “hpu”, “mps, “auto”) as well as custom accelerator instances.

  • device (int | str (default: 'auto')) – The device to use. Can be set to a non-negative index (int or str) or “auto” for automatic selection based on the chosen accelerator. If set to “auto” and accelerator is not determined to be “cpu”, then device will be set to the first available device.

  • prefix (str | None (default: None)) – Prefix of saved file names.

  • backup_url (str | None (default: None)) – URL to retrieve saved outputs from if not present on disk.

Returns:

Model with loaded state dictionaries.

Examples

>>> model = ModelClass.load(save_path, adata) # use the name of the model class used to save
>>> model.get_....
classmethod MULTIVI.load_query_data(adata, reference_model, inplace_subset_query_vars=False, accelerator='auto', device='auto', unfrozen=False, freeze_dropout=False, freeze_expression=True, freeze_decoder_first_layer=True, freeze_batchnorm_encoder=True, freeze_batchnorm_decoder=False, freeze_classifier=True)[source]#

Online update of a reference model with scArches algorithm [Lotfollahi et al., 2021].

Parameters:
  • adata (AnnData) – AnnData organized in the same way as data used to train model. It is not necessary to run setup_anndata, as AnnData is validated against the registry.

  • reference_model (Union[str, BaseModelClass]) – Either an already instantiated model of the same class, or a path to saved outputs for reference model.

  • inplace_subset_query_vars (bool (default: False)) – Whether to subset and rearrange query vars inplace based on vars used to train reference model.

  • accelerator (str (default: 'auto')) – Supports passing different accelerator types (“cpu”, “gpu”, “tpu”, “ipu”, “hpu”, “mps, “auto”) as well as custom accelerator instances.

  • device (Union[int, str] (default: 'auto')) – The device to use. Can be set to a non-negative index (int or str) or “auto” for automatic selection based on the chosen accelerator. If set to “auto” and accelerator is not determined to be “cpu”, then device will be set to the first available device.

  • unfrozen (bool (default: False)) – Override all other freeze options for a fully unfrozen model

  • freeze_dropout (bool (default: False)) – Whether to freeze dropout during training

  • freeze_expression (bool (default: True)) – Freeze neurons corersponding to expression in first layer

  • freeze_decoder_first_layer (bool (default: True)) – Freeze neurons corersponding to first layer in decoder

  • freeze_batchnorm_encoder (bool (default: True)) – Whether to freeze batchnorm weight and bias during training for encoder

  • freeze_batchnorm_decoder (bool (default: False)) – Whether to freeze batchnorm weight and bias during training for decoder

  • freeze_classifier (bool (default: True)) – Whether to freeze classifier completely. Only applies to SCANVI.

static MULTIVI.load_registry(dir_path, prefix=None)[source]#

Return the full registry saved with the model.

Parameters:
  • dir_path (str) – Path to saved outputs.

  • prefix (str | None (default: None)) – Prefix of saved file names.

Return type:

dict

Returns:

The full registry saved with the model

static MULTIVI.prepare_query_anndata(adata, reference_model, return_reference_var_names=False, inplace=True)[source]#

Prepare data for query integration.

This function will return a new AnnData object with padded zeros for missing features, as well as correctly sorted features.

Parameters:
  • adata (AnnData) – AnnData organized in the same way as data used to train model. It is not necessary to run setup_anndata, as AnnData is validated against the registry.

  • reference_model (Union[str, BaseModelClass]) – Either an already instantiated model of the same class, or a path to saved outputs for reference model.

  • return_reference_var_names (bool (default: False)) – Only load and return reference var names if True.

  • inplace (bool (default: True)) – Whether to subset and rearrange query vars inplace or return new AnnData.

Return type:

Union[AnnData, Index, None]

Returns:

Query adata ready to use in load_query_data unless return_reference_var_names in which case a pd.Index of reference var names is returned.

classmethod MULTIVI.register_manager(adata_manager)[source]#

Registers an AnnDataManager instance with this model class.

Stores the AnnDataManager reference in a class-specific manager store. Intended for use in the setup_anndata() class method followed up by retrieval of the AnnDataManager via the _get_most_recent_anndata_manager() method in the model init method.

Notes

Subsequent calls to this method with an AnnDataManager instance referring to the same underlying AnnData object will overwrite the reference to previous AnnDataManager.

MULTIVI.save(dir_path, prefix=None, overwrite=False, save_anndata=False, save_kwargs=None, **anndata_write_kwargs)[source]#

Save the state of the model.

Neither the trainer optimizer state nor the trainer history are saved. Model files are not expected to be reproducibly saved and loaded across versions until we reach version 1.0.

Parameters:
  • dir_path (str) – Path to a directory.

  • prefix (str | None (default: None)) – Prefix to prepend to saved file names.

  • overwrite (bool (default: False)) – Overwrite existing data or not. If False and directory already exists at dir_path, error will be raised.

  • save_anndata (bool (default: False)) – If True, also saves the anndata

  • save_kwargs (dict | None (default: None)) – Keyword arguments passed into save().

  • anndata_write_kwargs – Kwargs for write()

classmethod MULTIVI.setup_anndata(adata, layer=None, batch_key=None, size_factor_key=None, categorical_covariate_keys=None, continuous_covariate_keys=None, protein_expression_obsm_key=None, protein_names_uns_key=None, **kwargs)[source]#

Sets up the AnnData object for this model.

A mapping will be created between data fields used by this model to their respective locations in adata. None of the data in adata are modified. Only adds fields to adata.

Parameters:
  • adata (AnnData) – AnnData object. Rows represent cells, columns represent features.

  • layer (str | None (default: None)) – if not None, uses this as the key in adata.layers for raw count data.

  • batch_key (str | None (default: None)) – key in adata.obs for batch information. Categories will automatically be converted into integer categories and saved to adata.obs[‘_scvi_batch’]. If None, assigns the same batch to all the data.

  • size_factor_key (str | None (default: None)) – key in adata.obs for size factor information. Instead of using library size as a size factor, the provided size factor column will be used as offset in the mean of the likelihood. Assumed to be on linear scale.

  • categorical_covariate_keys (list[str] | None (default: None)) – keys in adata.obs that correspond to categorical data. These covariates can be added in addition to the batch covariate and are also treated as nuisance factors (i.e., the model tries to minimize their effects on the latent space). Thus, these should not be used for biologically-relevant factors that you do _not_ want to correct for.

  • continuous_covariate_keys (list[str] | None (default: None)) – keys in adata.obs that correspond to continuous data. These covariates can be added in addition to the batch covariate and are also treated as nuisance factors (i.e., the model tries to minimize their effects on the latent space). Thus, these should not be used for biologically-relevant factors that you do _not_ want to correct for.

  • protein_expression_obsm_key (str | None (default: None)) – key in adata.obsm for protein expression data.

  • protein_names_uns_key (str | None (default: None)) – key in adata.uns for protein names. If None, will use the column names of adata.obsm[protein_expression_obsm_key] if it is a DataFrame, else will assign sequential names to proteins.

MULTIVI.to_device(device)[source]#

Move model to device.

Parameters:

device (str | int) – Device to move model to. Options: ‘cpu’ for CPU, integer GPU index (eg. 0), or ‘cuda:X’ where X is the GPU index (eg. ‘cuda:0’). See torch.device for more info.

Examples

>>> adata = scvi.data.synthetic_iid()
>>> model = scvi.model.SCVI(adata)
>>> model.to_device('cpu')      # moves model to CPU
>>> model.to_device('cuda:0')   # moves model to GPU 0
>>> model.to_device(0)          # also moves model to GPU 0
MULTIVI.train(max_epochs=500, lr=0.0001, accelerator='auto', devices='auto', train_size=0.9, validation_size=None, shuffle_set_split=True, batch_size=128, weight_decay=0.001, eps=1e-08, early_stopping=True, save_best=True, check_val_every_n_epoch=None, n_steps_kl_warmup=None, n_epochs_kl_warmup=50, adversarial_mixing=True, datasplitter_kwargs=None, plan_kwargs=None, **kwargs)[source]#

Trains the model using amortized variational inference.

Parameters:
  • max_epochs (int (default: 500)) – Number of passes through the dataset.

  • lr (float (default: 0.0001)) – Learning rate for optimization.

  • accelerator (str (default: 'auto')) – Supports passing different accelerator types (“cpu”, “gpu”, “tpu”, “ipu”, “hpu”, “mps, “auto”) as well as custom accelerator instances.

  • devices (int | list[int] | str (default: 'auto')) – The devices to use. Can be set to a non-negative index (int or str), a sequence of device indices (list or comma-separated str), the value -1 to indicate all available devices, or “auto” for automatic selection based on the chosen accelerator. If set to “auto” and accelerator is not determined to be “cpu”, then devices will be set to the first available device.

  • train_size (float (default: 0.9)) – Size of training set in the range [0.0, 1.0].

  • validation_size (float | None (default: None)) – Size of the test set. If None, defaults to 1 - train_size. If train_size + validation_size < 1, the remaining cells belong to a test set.

  • shuffle_set_split (bool (default: True)) – Whether to shuffle indices before splitting. If False, the val, train, and test set are split in the sequential order of the data according to validation_size and train_size percentages.

  • batch_size (int (default: 128)) – Minibatch size to use during training.

  • weight_decay (float (default: 0.001)) – weight decay regularization term for optimization

  • eps (float (default: 1e-08)) – Optimizer eps

  • early_stopping (bool (default: True)) – Whether to perform early stopping with respect to the validation set.

  • save_best (bool (default: True)) – Save the best model state with respect to the validation loss, or use the final state in the training procedure

  • check_val_every_n_epoch (int | None (default: None)) – Check val every n train epochs. By default, val is not checked, unless early_stopping is True. If so, val is checked every epoch.

  • n_steps_kl_warmup (int | None (default: None)) – Number of training steps (minibatches) to scale weight on KL divergences from 0 to 1. Only activated when n_epochs_kl_warmup is set to None. If None, defaults to floor(0.75 * adata.n_obs).

  • n_epochs_kl_warmup (int | None (default: 50)) – Number of epochs to scale weight on KL divergences from 0 to 1. Overrides n_steps_kl_warmup when both are not None.

  • adversarial_mixing (bool (default: True)) – Whether to use adversarial training to penalize the model for umbalanced mixing of modalities.

  • datasplitter_kwargs (dict | None (default: None)) – Additional keyword arguments passed into DataSplitter.

  • plan_kwargs (dict | None (default: None)) – Keyword args for TrainingPlan. Keyword arguments passed to train() will overwrite values present in plan_kwargs, when appropriate.

  • **kwargs – Other keyword args for Trainer.

MULTIVI.view_anndata_setup(adata=None, hide_state_registries=False)[source]#

Print summary of the setup for the initial AnnData or a given AnnData object.

Parameters:
  • adata (Union[AnnData, MuData, None] (default: None)) – AnnData object setup with setup_anndata or transfer_fields().

  • hide_state_registries (bool (default: False)) – If True, prints a shortened summary without details of each state registry.

Return type:

None

static MULTIVI.view_setup_args(dir_path, prefix=None)[source]#

Print args used to setup a saved model.

Parameters:
  • dir_path (str) – Path to saved outputs.

  • prefix (str | None (default: None)) – Prefix of saved file names.

Return type:

None