`spateo.preprocessing.transform`#

Miscellaneous non-normalizing data transformations on AnnData objects

Module Contents#

`log1p`(X[, base, copy])	Computes the natural logarithm of the data matrix (unless different base is chosen using the base argument)
`log1p_sparse`(X, *[, base, copy])	Called if log1p is called with a sparse matrix input.
`log1p_array`(X, *[, base, copy])	Called if log1p is called with a numpy array input
`log1p_anndata`(→ Optional[anndata.AnnData])	Called if log1p is called with an AnnData input
`scale`(X[, zero_center, max_value, copy, layer, obsm, ...])	Scale variables to unit variance and optionally zero mean. Variables that are constant across all observations
`scale_sparse`(X, *[, zero_center, max_value, copy, ...])	Called if log1p is called with a sparse matrix input
`scale_array`(X, *[, zero_center, max_value, copy, ...])	Called if log1p is called with a numpy array input
`scale_anndata`(→ Optional[anndata.AnnData])	Called if scale is called with an AnnData object
`_get_mean_var`(X, *[, axis])	Wrapper for computing row-wise or column-wise mean+variance on sparse array-likes.
`sparse_mean_variance_axis`(mtx, axis)	Mean and variance of sparse array.
`sparse_mean_var_minor_axis`(data, indices, major_len, ...)	Given array for a csr matrix, returns means and variances for each column.
`sparse_mean_var_major_axis`(data, indices, indptr, ...)	Given array for a csc matrix, returns means and variances for each row.

spateo.preprocessing.transform.log1p(X: anndata.AnnData | numpy.ndarray | scipy.sparse.spmatrix, base: int | None = None, copy: bool = False)[source]#

Computes the natural logarithm of the data matrix (unless different base is chosen using the base argument)

Parameters:

X: Either full AnnData object or .X. Rows correspond to cells and columns to genes.
base: Natural log is used by default.
copy: If an AnnData is passed, determines whether a copy is returned.
layer: Layer to transform. If None, will transform .X. If given both argument to layer and obsm, argument to layer will take priority.
obsm: Entry in .obsm to transform. If None, will transform .X.

Returns:

If copy is True or input is numpy array/sparse matrix, returns updated data array. Otherwise, returns updated AnnData object.

spateo.preprocessing.transform.log1p_sparse(X, *, base: int | None = None, copy: bool = False)[source]#: Called if log1p is called with a sparse matrix input.

spateo.preprocessing.transform.log1p_array(X, *, base: int | None = None, copy: bool = False)[source]#: Called if log1p is called with a numpy array input

spateo.preprocessing.transform.log1p_anndata(adata, *, base: int | None = None, copy: bool = False, layer: str | None = None, obsm: str | None = None) → anndata.AnnData | None[source]#: Called if log1p is called with an AnnData input

spateo.preprocessing.transform.scale(X: anndata.AnnData | scipy.sparse.spmatrix | numpy.ndarray, zero_center: bool = True, max_value: float | None = None, copy: bool = False, layer: str | None = None, obsm: str | None = None, return_mean_std: bool = False)[source]#

Scale variables to unit variance and optionally zero mean. Variables that are constant across all observations will be set to 0.

Parameters:

X: Either full AnnData object or .X. Rows correspond to cells and columns to genes.
zero_center: If False, will not center variables.
max_value: Truncate to this value after scaling.
copy: If an AnnData is passed, determines whether a copy is returned.
layer: Layer to transform. If None, will transform .X. If given both argument to layer and obsm, argument to layer will take priority.
obsm: Entry in .obsm to transform. If None, will transform .X.
return_mean_std: Set True to return computed feature means and feature standard deviations.

Returns:

Depending on copy returns or updates adata with a scaled adata.X, annotated with ‘mean’ and ‘std’ in adata.var.

spateo.preprocessing.transform.scale_sparse(X, *, zero_center: bool = True, max_value: float | None = None, copy: bool = False, return_mean_std: bool = False)[source]#: Called if log1p is called with a sparse matrix input

spateo.preprocessing.transform.scale_array(X, *, zero_center: bool = True, max_value: float | None = None, copy: bool = False, return_mean_std: bool = False)[source]#: Called if log1p is called with a numpy array input

spateo.preprocessing.transform.scale_anndata(adata: anndata.AnnData, *, zero_center: bool = True, max_value: float | None = None, copy: bool = False, layer: str | None = None, obsm: str | None = None) → anndata.AnnData | None[source]#: Called if scale is called with an AnnData object

spateo.preprocessing.transform._get_mean_var(X, *, axis=0)[source]#: Wrapper for computing row-wise or column-wise mean+variance on sparse array-likes.

spateo.preprocessing.transform.sparse_mean_variance_axis(mtx: scipy.sparse.spmatrix, axis: int)[source]#

Mean and variance of sparse array.

Parameters:

mtx: scipy.sparse.csr_matrix or scipy.sparse.csc_matrix
axis: int Either 0 or 1. Determines which axis mean and variance are computed along

spateo.preprocessing.transform.sparse_mean_var_minor_axis(data, indices, major_len, minor_len, dtype)[source]#: Given array for a csr matrix, returns means and variances for each column.

spateo.preprocessing.transform.sparse_mean_var_major_axis(data, indices, indptr, major_len, minor_len, dtype)[source]#: Given array for a csc matrix, returns means and variances for each row.