sccoral.model.SCCORAL

sccoral.model.SCCORAL#

class sccoral.model.SCCORAL(adata, n_latent=10, alpha_l1=1000, n_hidden=128, n_layers=1, dropout_rate=0.1, dispersion='gene', log_variational=True, latent_distribution='ln', gene_likelihood='nb', use_batch_norm='both', use_layer_norm=False, use_observed_lib_size=True, **vae_kwargs)#

Single-cell COvariate-informed Regularized variational Autoencoder with Linear Decoder

Parameters:

  • adata (AnnData) – Registered AnnData object

  • n_latent (int (default: 10)) – Number of latent dimensions, approximated dimensionality of dataset

  • alpha_l1 (Tunable_[float] (default: 1000)) – Regularization strength in decoder

  • n_hidden (Tunable_[int] (default: 128)) – Number of hidden layers in encoder

  • n_layers (Tunable_[int] (default: 1)) – Number of layers in encoder neural network (see LSCVI)

  • dropout_rate (Tunable_[float] (default: 0.1)) – Dropout rate for neural networks (see LSCVI)

  • dispersion (Literal['gene', 'gene-batch', 'gene-cell'] (default: 'gene')) – Whether dispersion parameters of genes are fit on the level of 1) datasets 2) batches 3) cells (not implemented: labels)

  • log_variational (bool (default: True)) – Whether to log(x+1) counts x during encoding

  • latent_distribution (Literal['normal', 'ln'] (default: 'ln')) – Prior on latent space

  • gene_likelihood (Tunable_[Literal['nb', 'zinb', 'poisson']] (default: 'nb')) –

    One of (see scVI/LSCVI)

    • nb - Negative binomial distribution

    • zinb - Zero inflated negative binomial distribution

    • poisson - Poisson distribution

  • use_batch_norm (Literal['encoder', 'decoder', 'both', 'none'] (default: 'both')) – Batch norm in encoder/decoder

  • use_layer_norm (bool (default: False)) – Layer norm in encoder

  • **model_kwargs – Keyword arguments for _module

Examples

>>> adata = sccoral.data.simulation_dataset()
>>> sccoral.model.setup_anndata(adata,
                                categorical_covariate='categorical_covariate',
                                continuous_covariate='continuous_covariate'
                                )
>>> m = sccoral.model(adata, n_latent=7)
>>> m.train()
>>> representation = m.get_latent_representation()  # pd.DataFrame cells x n_latent
>>> loadings = m.get_loadings()  # pd.DataFrame genes x n_latent
>>> r2 = m.get_explained_variance_per_factor()  # pd.DataFrame 1 x n_latent

Notes

Upcoming documentation 1. :doc:

References

[].

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.

get_anndata_manager(adata[, required])

Retrieves the AnnDataManager for a given AnnData object.

get_elbo([adata, indices, batch_size])

Return the ELBO for the data.

get_explained_variance_per_factor(adata[, ...])

Compute explained variance per factor

get_from_registry(adata, registry_key)

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

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

Get latent representation of cells in anndata object

get_loadings([set_column_names])

Extract linear weights of decoder

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

Return the marginal LL for the data.

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

Return the reconstruction error for the data.

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

Instantiate a model from the saved output.

load_registry(dir_path[, prefix])

Return the full registry saved with the model.

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[, batch_key, ...])

Sets up the AnnData object for this model.

to_device(device)

Move model to device.

train([max_epochs, pretraining, use_gpu, ...])

Train sccoral model

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#

SCCORAL.adata#

Data attached to model instance.

SCCORAL.adata_manager#

Manager instance associated with self.adata.

SCCORAL.device#

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

SCCORAL.history#

Returns computed metrics during training.

SCCORAL.is_trained#

Whether the model has been trained.

SCCORAL.summary_string#

Summary string of the model.

SCCORAL.test_indices#

Observations that are in test set.

SCCORAL.train_indices#

Observations that are in train set.

SCCORAL.validation_indices#

Observations that are in validation set.

Methods#

classmethod SCCORAL.convert_legacy_save(dir_path, output_dir_path, overwrite=False, prefix=None, **save_kwargs)#

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

SCCORAL.deregister_manager(adata=None)#

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.

SCCORAL.get_anndata_manager(adata, required=False)#

Retrieves the AnnDataManager for a given AnnData object.

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

SCCORAL.get_elbo(adata=None, indices=None, batch_size=None)#

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

SCCORAL.get_explained_variance_per_factor(adata, set_column_names=False)#

Compute explained variance per factor

Parameters:

  • adata (None | AnnData) – AnnData object to embed. If None use stored anndata.AnnData

  • set_column_names (bool (default: False)) – Whether to set the column names to covariate names

Return type:

DataFrame

Returns:

Pandas DataFrame 1 x n_latent

SCCORAL.get_from_registry(adata, registry_key)#

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.

SCCORAL.get_latent_representation(adata=None, indices=None, give_mean=True, mc_samples=5000, batch_size=None, return_dist=False, set_column_names=True, suffix=None)#

Get latent representation of cells in anndata object

Parameters:

  • adata (AnnData | None (default: None)) – AnnData object to embed. If None use stored anndata.AnnData

  • indices (Iterable[int] | None (default: None)) – Indices of cells to retrieve (see scvi-tools)

  • give_mean (bool (default: True)) – Whether to give the full distribution or mean of distribution. Defaults to mean See scvi-tools

  • mc_samples (int (default: 5000)) – For distributions with no closed analytical solution - how many samples to draw (see scvi-tools)

  • batch_size (int | None (default: None)) – Batch size during inference.

  • return_dist (bool (default: False)) – Whether to return single-measurement values (False) or parameters of the distribution (True) See scvi-tools

  • set_column_names (bool (default: True)) – Whether to set the column names to covariate names (defaults to True)

  • suffix (str | None (default: None)) – Whether to add a suffix (e.g. __factor) so that columns in dataframe are better distinguishable from metadata info. Per default, no suffix is added.

Return type:

DataFrame | tuple[ndarray, ndarray]

Returns:

Pandas DataFrame n_cells x n_latent

SCCORAL.get_loadings(set_column_names=True)#

Extract linear weights of decoder

Parameters:

set_column_names (bool (default: True)) – Whether to set the column names to covariate names

Return type:

DataFrame

Returns:

Pandas DataFrame n_genes x n_latent

SCCORAL.get_marginal_ll(adata=None, indices=None, n_mc_samples=1000, batch_size=None, return_mean=True, **kwargs)#

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]

SCCORAL.get_reconstruction_error(adata=None, indices=None, batch_size=None)#

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

classmethod SCCORAL.load(dir_path, adata=None, accelerator='auto', device='auto', prefix=None, backup_url=None)#

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)
>>> model.get_....
static SCCORAL.load_registry(dir_path, prefix=None)#

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

classmethod SCCORAL.register_manager(adata_manager)#

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.

SCCORAL.save(dir_path, prefix=None, overwrite=False, save_anndata=False, save_kwargs=None, **anndata_write_kwargs)#

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 SCCORAL.setup_anndata(adata, batch_key=None, categorical_covariates=None, continuous_covariates=None, layer=None, **kwargs)#

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.

Each model class deriving from this class provides parameters to this method according to its needs. To operate correctly with the model initialization, the implementation must call register_manager() on a model-specific instance of AnnDataManager.

SCCORAL.to_device(device)#

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
SCCORAL.train(max_epochs=500, pretraining=True, use_gpu=True, accelerator='auto', devices='auto', validation_size=0.1, batch_size=128, early_stopping=True, pretraining_max_epochs=500, pretraining_early_stopping=True, pretraining_early_stopping_metric='reconstruction_loss_train', pretraining_min_delta=0.0, pretraining_early_stopping_patience=5, plan_kwargs=None, trainer_kwargs=None, **kwargs)#

Train sccoral model

Training is split into pretraining (only training on covariates, frozen z_encoder weights) and training (unfrozen weights). Same training procedure as for scVI/LSCVI except for pretraining.

Parameters:

  • max_epochs (int (default: 500)) – Maximum epochs during training

  • max_pretraining_epochs – Maximum epochs during pretraining. If None, same as max_epochs

  • accelerator (Optional[Literal['cpu', 'gpu', 'auto']] (default: 'auto')) – cpu/gpu/auto: auto automatically detects available devices

  • devices (default: 'auto') – If auto, automatically detects available devices

  • validation_size (None | float (default: 0.1)) – Size of validation split (0-1). Rest is train split

  • batch_size (int (default: 128)) – Size of minibatches during training

  • early_stopping (Tunable_[bool] (default: True)) – Enable early stopping during training

  • pretraining (Tunable_[bool] (default: True)) – Whether to conduct pretraining

  • pretraining_max_epochs (Tunable_[int] (default: 500)) – Maximum number of epochs for pretraining to continue.

  • pretraining_early_stopping (Tunable_[bool] (default: True)) – Enable early stopping during pretraining

  • plan_kwargs (None | dict[str, Any] (default: None)) – Training keyword arguments passed to sccoral.train.TrainingPlan

  • trainer_kwargs (None | dict[str, Any] (default: None)) – Additional keyword arguments passed to scvi.train.TrainRunner

  • kwargs – Not passed.

Return type:

None

Returns:

Training runner (scvi-tools wrapper of pytorch lightning trainer.)

SCCORAL.view_anndata_setup(adata=None, hide_state_registries=False)#

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 SCCORAL.view_setup_args(dir_path, prefix=None)#

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

Contents