smoother.models.reduction._spvae

Classes

`SpatialLoss`	Spatial loss.
`SPVAE`	Add spatial loss to the latent representation in the VAE model.
`SpatialVAE`	Spatially-aware Variational Autoencoder model.

Functions

set_params_online_update(module, unfrozen, ...)

Freeze parts of network for scArches.

Module Contents

class smoother.models.reduction._spvae.SpatialLoss(prior, spatial_weights: List[smoother.weights.SpatialWeightMatrix] = None, rho=1, use_sparse=True, standardize_cov=False)

Bases: torch.nn.Module

Spatial loss.

The spatial smoothing loss on a spatial random variable (num_group x num_spot).

prior

The prior spatial process model, can be one of ‘sma’,’sar’, ‘car’, ‘icar’.

Type:: str

spatial_weights

Spatial weight matrix collection of length num_group or 1. If 1 then all groups will be subject to the same covariance.

Type:: List[SpatialWeightMatrix]

rho

The spatial autocorrelation parameter (for SpatialWeightMatrix.get_inv_cov).

Type:: float

use_sparse

Whether to use sparse inverse covariance matrix in the calculation.

Type:: bool

standardize_cov

Whether to standardize the covariance matrix to have same variance (1) across locations. Only proper covariance can be standardized.

Type:: bool

inv_cov

Inverse covariance (precision) matrix of spatial variables of each group, (num_group or 1) x n x n. If the first dimension has length 1, all groups will have the same covariance structure.

Type:: 3D tensor

inv_cov_2d_sp

Sparse block diagonal inverse covariance (precision) matrix.

Type:: 2D sparse tensor

confidences

Relative prior confidence of each group. The higher the confidence, the stronger the smoothing will be. If float, all groups will have the same confidence.

Type:: 1D tensor or float

_sanity_check() → None: Check whether the spatial loss is defined properly.

estimate_confidence(ref_exp, st_exp, method='lr') → None

Estimate the relative confidence for each group.

The covariance matrix will be scaled accordingly.

Parameters:

ref_exp (2D tensor) – Bulk expression signiture matrix, num_gene x num_group.
st_exp (2D tensor) – Spatial expression matrix, num_gene x num_spot.
method (str) – Method used to estimate variance.

forward(coefs, normalize=True)

Calculate spatial loss.

Parameters:: coefs (2D tensor) – Columns of regression coefficients, num_group x num_spot.

calc_corr_decay_stats(coords, min_k=0, max_k=50, cov_ind=0, return_var=False)

Calculate spatial covariance decay over degree of neighborhoods.

Covariance measured between k-nearest neighbors. If the number of covariance matrices (i.e. self.inv_cov.shape[0]) is larger than 1, use ‘cov_ind’ to select the covariance matrix to use.

Parameters:

coords (2D tensor) – Coordinates of spots, num_spot x 2.
min_k – Minimum number of k in k-nearest neighbors. k = 0: self.
max_k – Maximum number of k in k-nearest neighbors.
cov_ind (int) – Index of the covariance matrix to use.
return_var (bool) – Whether to return variance stats.

Returns:

Correlation decay quantiles. var_quantiles_df (pd.DataFrame): Per-spot variance quantiles.

Return type:

corr_decay_quantiles_df (pd.DataFrame)

smoother.models.reduction._spvae.set_params_online_update(module, unfrozen, freeze_decoder_first_layer, freeze_batchnorm_encoder, freeze_batchnorm_decoder, freeze_dropout, freeze_expression, freeze_classifier): Freeze parts of network for scArches.

class smoother.models.reduction._spvae.SPVAE(n_input: int, spatial_loss: smoother.SpatialLoss | None = None, lambda_spatial_loss=0.1, sp_loss_on: Literal['z', 'mean_z'] = 'mean_z', n_hidden: int = 128, n_latent: int = 10, n_layers: int = 1, dropout_rate: float = 0.1, dispersion: Literal['gene', 'gene-batch', 'gene-label', 'gene-cell'] = 'gene', gene_likelihood: Literal['zinb', 'nb', 'poisson'] = 'nb', latent_distribution: Literal['normal', 'ln'] = 'normal', **model_kwargs)

Bases: scvi.module.VAE

Add spatial loss to the latent representation in the VAE model.

Parameters:

spatial_loss – Spatial loss to apply on the latent representation. If None, no spatial loss.
lambda_spatial_loss – Weight of the spatial loss.
sp_loss_on – Whether to apply the spatial loss on the sampled latent representation (‘z’) or on the encoded mean of the latent distribution (‘mean_z’).

loss(tensors, inference_outputs, generative_outputs, kl_weight: float = 1.0)

classmethod from_vae(vae_module: scvi.module.VAE, spatial_loss: smoother.SpatialLoss | None = None, lambda_spatial_loss=0.1, sp_loss_on: Literal['z', 'mean_z'] = 'mean_z')

class smoother.models.reduction._spvae.SpatialVAE(st_adata: anndata.AnnData, spatial_loss: smoother.SpatialLoss | None = None, lambda_spatial_loss=0.1, sp_loss_on: Literal['z', 'mean_z'] = 'mean_z', n_hidden: int = 128, n_latent: int = 10, n_layers: int = 1, dropout_rate: float = 0.0, dispersion: Literal['gene', 'gene-batch', 'gene-label', 'gene-cell'] = 'gene', gene_likelihood: Literal['zinb', 'nb', 'poisson'] = 'nb', latent_distribution: Literal['normal', 'ln'] = 'normal', **model_kwargs)

Bases: scvi.model.SCVI

Spatially-aware Variational Autoencoder model.

_data_splitter_cls

_module_cls

train(max_epochs: int = 400, lr: float = 0.01, accelerator: str = 'auto', devices: int | List[int] | str = 'auto', plan_kwargs: dict | None = None, **kwargs): Trains the model without mini-batch.

classmethod from_rna_model(st_adata: anndata.AnnData, sc_model: scvi.model.SCVI, spatial_loss: smoother.SpatialLoss | None = None, lambda_spatial_loss=0.1, sp_loss_on: Literal['z', 'mean_z'] = 'mean_z', unfrozen: bool = False, freeze_dropout: bool = False, freeze_expression: bool = True, freeze_decoder_first_layer: bool = True, freeze_batchnorm_encoder: bool = True, freeze_batchnorm_decoder: bool = False, freeze_classifier: bool = True, **spvae_kwargs)

Alternate constructor for exploiting a pre-trained model on RNA-seq data.

Note that because of the dropout layer, even though the new instance is initialized with the same parameters as the pre-trained model, new_instance.get_latent_representation() may not return the same latent representation as the pre-trained model.

Parameters:

st_adata – registed anndata object
sc_model – pretrained SCVI model