omicverse.bulk2single.Single2Spatial

omicverse.bulk2single.Single2Spatial#

class omicverse.bulk2single.Single2Spatial(single_data, spatial_data, celltype_key, spot_key=['xcoord', 'ycoord'], top_marker_num=500, marker_used=True, gpu=0)[source]#

Deep-learning mapper that projects single-cell profiles onto spatial coordinates.

Parameters:

  • single_data (anndata.AnnData) – Single-cell reference AnnData containing expression and cell types.

  • spatial_data (anndata.AnnData) – Spatial transcriptomics AnnData used as mapping target.

  • celltype_key (str) – Column name in single_data.obs containing cell-type labels.

  • spot_key (list) – Column names in spatial_data.obs storing x/y coordinates.

  • top_marker_num (int) – Number of marker genes used to train mapping model.

  • marker_used (bool) – Whether to restrict training features to marker genes.

  • gpu (Union[int,str]) – Compute device selector (CUDA index, 'mps', or CPU fallback).

Returns:

Initializes single-cell to spatial mapping workflow.

Return type:

None

Examples

>>> st_model = ov.bulk2single.Single2Spatial(single_data=single_data, spatial_data=st_data, celltype_key="Cell_type")
__init__(single_data, spatial_data, celltype_key, spot_key=['xcoord', 'ycoord'], top_marker_num=500, marker_used=True, gpu=0)[source]#

Initialize Single2Spatial model for mapping single cells to spatial coordinates.

Parameters:

  • single_data (anndata.AnnData) – Single-cell reference AnnData for source expression profiles.

  • spatial_data (anndata.AnnData) – Spatial reference AnnData providing coordinates and spot expression.

  • celltype_key (str) – Name of the cell-type annotation column in single_data.obs.

  • spot_key (list) – Two-column key in spatial_data.obs defining x/y coordinates.

  • top_marker_num (int) – Number of marker genes selected for mapping.

  • marker_used (bool) – Whether marker-based feature selection is enabled.

  • gpu (Union[int,str]) – Device selector for model training.

Returns:

None

Methods

__init__(single_data, spatial_data, celltype_key)

Initialize Single2Spatial model for mapping single cells to spatial coordinates.

load(modelsize[, df_load_dir, ...])

Load a pre-trained Single2Spatial model and perform mapping.

save([df_save_dir, df_save_name])

Save the trained Single2Spatial model.

spot_assess()

Assess and aggregate predicted spatial data at the spot level.

train(spot_num, cell_num[, df_save_dir, ...])

Train the deep neural network for single-cell to spatial mapping.
Contents