Spatial architecture - niche composition

Spatial composition analysis identifies multi-cellular niches that composed by cell types with specific proportion.

You can download adata_combine_7celltype.h5ad from https://zenodo.org/records/14588408.

MIBI-TOF TNBC samples were used as in tutorial: Cell type proximity

Read

import warnings
warnings.filterwarnings("ignore")
import scanpy as sc
import SOAPy_st as sp
import numpy as np

To simplify the analyzed data, the cell types were combined, and DK cells, NK cells, Neutrophil cells, Treg cells, and other immune cell categories were collectively referred to as other.

adata_com = sc.read('/csb2/project/SpatialPackage_whq/Tutorial/data/adata_combine_7celltype.h5ad')

Constructing spatial network

the cell-cell network was built using 100 pixels as the radius to capture the composition pattern of more surrounding cells

sp.pp.make_network(adata_com, sample_key='sample', method='radius', cutoff=100, scale=1.0, cluster_key='cluster')

WARNING:root:adata has not been initialized, and adata.uns['SOAPy'] has been established

AnnData object with n_obs × n_vars = 211649 × 44
    obs: 'sample', 'leiden', 'cluster'
    var: 'highly_variable'
    uns: 'cluster_colors', 'leiden', 'leiden_colors', 'neighbors', 'pca', 'umap', 'var_for_clustering', 'SOAPy'
    obsm: 'X_pca', 'X_umap', 'spatial'
    varm: 'PCs'
    obsp: 'connectivities', 'distances'

C-niche cluster

Each cell was represented by the proportion of cell type of its surrounding cells (niche). 211649 cells from 41 TNBC samples were clustered into 30 C-niche clusters.

sp.tl.get_c_niche(adata=adata_com, k_max=30, celltype_key='cluster', sample_key='sample')

AnnData object with n_obs × n_vars = 211649 × 44
    obs: 'sample', 'leiden', 'cluster', 'C_niche'
    var: 'highly_variable'
    uns: 'cluster_colors', 'leiden', 'leiden_colors', 'neighbors', 'pca', 'umap', 'var_for_clustering', 'SOAPy'
    obsm: 'X_pca', 'X_umap', 'spatial'
    varm: 'PCs'
    obsp: 'connectivities', 'distances'

Show the cellular composition of each C-niche.

fig = sp.pl.show_celltype_niche_heatmap(adata=adata_com, cmap = 'coolwarm', figsize=(20, 16))

Show the cell type composition of each sample.

fig = sp.pl.show_celltype_sample_heatmap(adata=adata_com, cmap = 'coolwarm', figsize=(20, 16))

Show the C-niche composition of each sample.

fig = sp.pl.show_niche_sample_heatmap(adata=adata_com, cmap = 'coolwarm', figsize=(20, 16))

Show the composition of the microenvironment belonging to C-niche5 cells in sample 1

# Fixed color map of each cell type
color_map = {
    'other': '#999999',
    'B': '#765005',
    'CD4 T': '#0780cf',
    'CD8 T': '#fa6d1d',
    'Endothelial cell': '#024b51',
    'Epithelial cell': '#b6b51f',
    'Macrophage': '#da1f18',
    'Mesenchymal cell': '#701866'
}

# Black boundary: cells of C-niche 2, Gray boundary: cells of C-niche
# Filled areas with colors of cell types: the range of microenvironments labeled as C-niche 2 cells.

fig = sp.pl.show_niche_environment(
    adata=adata_com,
    celltype_key='cluster',
    sample_key='sample',
    sample_id=1,
    color=color_map,
    niche_key='C_niche',
    niche=5
)

Tensor decomposition of niche

In order to investigate the combinational effects of cell types and niches on patient heterogeneity, the “Niche-CellType (without epithelial)-Sample” tensor (size: 30*7*41) was factorized to four factors

To highlight the more diverse microenvironment composition, epithelial cells were removed when constructing the tensor.

niche_tensor = sp.tl.TensorDecomposition()
adata_copy = adata_com[adata_com.obs['cluster'] != 'Epithelial cell', :].copy()
niche_tensor.tensor_with_obs(adata_copy, obs_factor=['sample', 'cluster', 'C_niche'])

Each sample is internally normalized followed by non-negative CP decomposition

niche_tensor.normalization(factor_name='sample')
weights, factors, nre = niche_tensor.CP_decomposition(rank=4, non_negative=True)

The sample factor matrix is clustered, and the samples with the similar microenvironment are clustered into one class.

sp.pl.show_factor_matrix_in_CP_tensor(tensor=niche_tensor, factor_name='sample', figsize=(8, 12))

Through the orthographic observation of the samples, the samples enriched in factor 3 have a large number of enriched regions of B cells, which may correspond to tertiary lymph node structures.

sp.pl.show_voronoi(adata_com, sample_id = 35, cluster_key='cluster', title = 'sample 35')

sp.pl.show_voronoi(adata_com, sample_id = 1, cluster_key='cluster', title = 'sample 1')