neighNMF

neighNMF

neighNMF(
    data,
    xCoordinate="X_centroid",
    yCoordinate="Y_centroid",
    zCoordinate=None,
    phenotype="phenotype",
    method="radius",
    radius=30,
    knn=10,
    imageId="imageid",
    subset=None,
    verbose=True,
    label="neighNMF",
    maxWorkers=None,
    outputDir=None,
    streamData=False,
    inferenceMode="phenotype_driven",
    layer=None,
    subsetMarker=None,
    nmfComponents=5,
    nmfRandomState=0,
    sdataTable=None,
)

Non-negative matrix factorisation of neighbourhood composition.

Builds the same cell x category neighbourhood matrix as sp.tl.neighCount, then factorises it into nmfComponents parts. Each cell gets a weight per component, and each component is a recognisable mixture of cell types — so a cell can sit partly in two neighbourhoods, which k-means cannot express.

Only the factorisation is kept: uns[label] holds the cell factors and uns[label + "_nmf_components"] the component definitions. The full neighbourhood matrix is discarded, which keeps the object small on big slides — run sp.tl.neighCount as well if you want it.

Parameters:

Name	Type	Description	Default
data	(AnnData | SpatialData | str, required)	The cell table. An AnnData, a path to an .h5ad file, a SpatialData object, or a path to a .zarr SpatialData store.	required
xCoordinate	str	Column in obs holding x positions.	'X_centroid'
yCoordinate	str	Column in obs holding y positions.	'Y_centroid'
zCoordinate	str	Column in obs holding z positions. Leave as None for 2D data; set it and the neighbourhood search becomes three-dimensional.	None
phenotype	str	Column in obs holding the cell type labels.	'phenotype'
method	str	How neighbourhoods are defined. 'radius' takes every cell within radius pixels; 'knn' takes the knn nearest cells regardless of distance.	'radius'
radius	float	Neighbourhood radius in the units of the coordinate columns, normally pixels. Used when method='radius'.	30
knn	int	Number of nearest neighbours. Used when method='knn'.	10
imageId	str	Column in obs holding image identifiers. Neighbourhoods never cross images.	'imageid'
subset	str	Process only this image.	None
label	str	Key in uns the cell factors are written to. label + "_nmf_components" holds the components.	'neighNMF'
maxWorkers	int	Cap on parallel workers. Defaults to max(1, cpu_count() - 1). Lower it to cut peak memory.	None
inferenceMode	str	'phenotype_driven' builds the neighbourhood matrix from phenotype labels; 'marker_driven' builds it from marker expression, reading the matrix selected by layer and binarising it if needed.	'phenotype_driven'
layer	str	Which matrix to read for inferenceMode='marker_driven'. None uses .X.	None
subsetMarker	list	Restrict 'marker_driven' counting to these markers.	None
nmfComponents	int	Number of components to factor into. The main thing to tune.	5
nmfRandomState	int	Seed for the NMF initialisation.	0
verbose	bool	Print progress messages.	True
outputDir	str	Directory to write the updated object to. With None the object is returned instead.	None
streamData	bool	Run out of core against an .h5ad path with CAP-AnnData, reading only the sections this function needs and writing back only what it changes. Requires data to be a path; the file is updated in place and None is returned.	False
sdataTable	str	Which SpatialData table to work on. Ignored for AnnData input, and optional when the store has exactly one table.	None

Returns:

Name	Type	Description
adata	AnnData | SpatialData | None	The updated object with uns[label] and uns[label + "_nmf_components"]. None in streaming mode.

Example

adata = sp.tl.neighNMF(
    adata, phenotype="phenotype", radius=30, nmfComponents=6
)

# What is each component made of?
adata.uns["neighNMF_nmf_components"]

# And where does each cell sit?
adata.uns["neighNMF"].head()