schist.inference

Functions

`fit_model`(→ Optional[anndata.AnnData])	Cluster cells using the nested Stochastic Block Model [Peixoto14],
`fit_model_multi`(→ [Union[List[anndata.AnnData]], ...)	Cluster cells into subgroups using multiple modalities.
`nested_model`(→ Optional[anndata.AnnData])	This function has been deprecated and it soon will be removed.
`flat_model`(→ Optional[anndata.AnnData])	This function has been deprecated and it soon will be removed.
`nested_model_multi`(→ Optional[List[anndata.AnnData]])	This function has been deprecated and it soon will be removed.

Package Contents

schist.inference.fit_model(adata: anndata.AnnData, deg_corr: bool = True, tolerance: float = 0.0001, n_sweep: int = 10, beta: float = np.inf, n_init: int = 100, model: Literal['nsbm', 'sbm', 'ppbm'] = 'nsbm', max_iter: int = 1000, collect_marginals: bool = True, refine_model: bool = False, refine_iter: int = 100, n_jobs: int = -1, key_added: str | None = None, adjacency: scipy.sparse.spmatrix | None = None, neighbors_key: str | None = 'neighbors', directed: bool = False, use_weights: bool = False, save_model: str | None = None, copy: bool = False, random_seed: int | None = None, dispatch_backend: str | None = 'loky') → anndata.AnnData | None

Cluster cells using the nested Stochastic Block Model [Peixoto14], performing Bayesian inference on node groups.

This requires having ran neighbors() or bbknn() first.

Parameters

adata: The annotated data matrix.
deg_corr: Whether to use degree correction in the minimization step. In many real world networks this is the case, although this doesn’t seem the case for KNN graphs used in scanpy.
tolerance: Tolerance for fast model convergence.
n_sweep: Number of iterations to be performed in the fast model MCMC greedy approach
beta: Inverse temperature for MCMC greedy approach
n_init: Number of concurrent minimizations to be performed. The final model will be a consensus over these models.
model: The SBM model to use. nsbm implements Nested Stochastic Block Model. sbm is the Stochastic Block Model. ppbm is the Planted Partition Block Model which only has an assortativity prior.
max_iter: Maximum number of iterations during minimization, set to infinite to stop minimization only on tolerance
collect_marginals: Collect marginal distribution of cells, that is the probability to belong to any cluster
refine_model: Wether to perform a further mcmc step to refine the model
refine_iter: Number of refinement iterations.
n_jobs: Number of parallel computations used during model initialization
key_added: adata.obs key under which to add the cluster labels.
adjacency: Sparse adjacency matrix of the graph, defaults to adata.uns[‘neighbors’][‘connectivities’] in case of scanpy<=1.4.6 or adata.obsp[neighbors_key][connectivity_key] for scanpy>1.4.6
neighbors_key: The key passed to sc.pp.neighbors
directed: Whether to treat the graph as directed or undirected.
use_weights: If True, edge weights from the graph are used in the computation (placing more emphasis on stronger edges). Note that this increases computation times
save_model: If provided, this will be the filename for the PartitionModeState to be saved. The PartitionModeState contains all the models minimized during inference.
copy: Whether to copy adata or modify it inplace.
random_seed: Random number to be used as seed for graph-tool

Returns

adata.obs[key_added]: Array of dim (number of cells) that stores the subgroup id (‘0’, ‘1’, …) for each cell.
adata.uns[‘schist’][model][‘stats’]: A dict with entropy and modularity values
adata.uns[‘schist’][model][‘params’]: A dict with the values for the parameters used
adata.obsm[‘CM_nsbm_level_{n}’] or adata.obsm[‘CM_model’]: A np.ndarray with cell probability of belonging to a specific group
adata.uns[‘schist’][model][‘state’]: The block model, to be used in case a gt state should be initialized

schist.inference.fit_model_multi(mdata: List[anndata.AnnData] | mudata.MuData, deg_corr: bool = True, tolerance: float = 0.0001, n_sweep: int = 10, beta: float = np.inf, n_init: int = 100, model: Literal['nsbm', 'sbm'] = 'nsbm', max_iter: int = 1000, collect_marginals: bool = True, refine_model: bool = False, refine_iter: int = 100, n_jobs: int = -1, overlap: bool = False, key_added: str | None = None, adjacency: List[scipy.sparse.spmatrix] | None = None, neighbors_key: List[str] | None = ['neighbors'], directed: bool = False, use_weights: bool = False, save_model: str | None = None, copy: bool = False, dispatch_backend: str | None = 'loky', random_seed: int | None = None) → [List[anndata.AnnData], mudata.MuData, None]

Cluster cells into subgroups using multiple modalities.

Cluster cells using the nested Stochastic Block Model [Peixoto14], performing Bayesian inference on node groups. This function takes multiple experiments, possibly across different modalities, and perform joint clustering.

This requires having ran neighbors() or bbknn() first. It also requires cells having the same names if coming from paired experiments

Parameters

mdata: A list of processed AnnData. Neighbors must have been already calculated. If a MuData object is passed, a model on the layered graph will be fitted. If you want to fit a model on the shared graph representation, e.g. WNN graph or a graph built on MOFA latent factors, you still can use the standard scs.inference.model() function.
deg_corr: Whether to use degree correction in the minimization step. In many real world networks this is the case, although this doesn’t seem the case for KNN graphs used in scanpy.
tolerance: Tolerance for fast model convergence.
n_sweep: Number of iterations to be performed in the fast model MCMC greedy approach
beta: Inverse temperature for MCMC greedy approach
n_init: Number of initial minimizations to be performed. The one with smaller entropy is chosen
refine_model: Wether to perform a further mcmc step to refine the model
refine_iter: Number of refinement iterations.
max_iter: Maximum number of iterations during minimization, set to infinite to stop minimization only on tolerance
overlap: Whether the different layers are dependent (overlap=True) or not (overlap=False)
n_jobs: Number of parallel computations used during model initialization
key_added: adata.obs key under which to add the cluster labels.
adjacency: Sparse adjacency matrix of the graph, defaults to adata.uns[‘neighbors’][‘connectivities’] in case of scanpy<=1.4.6 or adata.obsp[neighbors_key][connectivity_key] for scanpy>1.4.6
neighbors_key: The key passed to sc.pp.neighbors. If all AnnData share the same key, one only has to be specified, otherwise the full tuple of all keys must be provided
directed: Whether to treat the graph as directed or undirected.
use_weights: If True, edge weights from the graph are used in the computation (placing more emphasis on stronger edges). Note that this increases computation times
save_model: If provided, this will be the filename for the PartitionModeState to be saved
copy: Whether to copy adata or modify it inplace.
random_seed: Random number to be used as seed for graph-tool

Returns

adata.obs[key_added]: Array of dim (number of cells) that stores the subgroup id (‘0’, ‘1’, …) for each cell.
adata.uns[‘schist’][‘multi_level_params’]: A dict with the values for the parameters resolution, random_state, and n_iterations.
adata.uns[‘schist’][‘multi_level_stats’]: A dict with the values returned by mcmc_sweep
adata.obsm[‘CA_multi_nsbm_level_{n}’]: A np.ndarray with cell probability of belonging to a specific group
adata.uns[‘schist’][‘multi_level_state’]: The NestedBlockModel state object

schist.inference.nested_model(adata: anndata.AnnData, deg_corr: bool = True, tolerance: float = 1e-06, n_sweep: int = 10, beta: float = np.inf, n_init: int = 100, collect_marginals: bool = True, n_jobs: int = -1, refine_model: bool = False, refine_iter: int = 100, max_iter: int = 100000, *, restrict_to: Tuple[str, Sequence[str]] | None = None, random_seed: int | None = None, key_added: str = 'nsbm', adjacency: scipy.sparse.spmatrix | None = None, neighbors_key: str | None = 'neighbors', directed: bool = False, use_weights: bool = False, save_model: str | None = None, copy: bool = False, dispatch_backend: str | None = 'threads') → anndata.AnnData | None: This function has been deprecated and it soon will be removed. It now wraps scs.inference.fit_model() function.

schist.inference.flat_model(adata: anndata.AnnData, n_sweep: int = 10, beta: float = np.inf, tolerance: float = 1e-06, collect_marginals: bool = True, deg_corr: bool = True, n_init: int = 100, n_jobs: int = -1, refine_model: bool = False, refine_iter: int = 100, max_iter: int = 100000, *, restrict_to: Tuple[str, Sequence[str]] | None = None, random_seed: int | None = None, key_added: str = 'sbm', adjacency: scipy.sparse.spmatrix | None = None, neighbors_key: str | None = 'neighbors', directed: bool = False, use_weights: bool = False, save_model: str | None = None, copy: bool = False, dispatch_backend: str | None = 'threads') → anndata.AnnData | None: This function has been deprecated and it soon will be removed. It now wraps scs.inference.fit_model() function.

schist.inference.nested_model_multi(adatas: List[anndata.AnnData], deg_corr: bool = True, tolerance: float = 1e-06, n_sweep: int = 10, beta: float = np.inf, n_init: int = 100, collect_marginals: bool = True, n_jobs: int = -1, refine_model: bool = False, refine_iter: int = 100, overlap: bool = False, max_iter: int = 100000, *, random_seed: int | None = None, key_added: str = 'multi_nsbm', adjacency: List[scipy.sparse.spmatrix] | None = None, neighbors_key: List[str] | None = ['neighbors'], directed: bool = False, use_weights: bool = False, save_model: str | None = None, copy: bool = False, dispatch_backend: str | None = 'threads') → List[anndata.AnnData] | None: This function has been deprecated and it soon will be removed. It now wraps scs.inference.fit_model() function.