Tutorial 5: Nanostring Dataset

In this tutorial, we show using the model on 20 slices of nanostring. Relevant data can be obtained from https://drive.google.com/drive/folders/141__9Q4zYK_6A4stKsOE5WikIHgxFkhk

Import the relevant python analysis package

import scanpy as sc
import pandas as pd
from sklearn import metrics
import torch
import scipy
import matplotlib.pyplot as plt
import os
import warnings
warnings.filterwarnings('ignore')
import stCAMBL
import os
#Please change this path to your local R environment path
os.environ['R_HOME'] = '/data3/wkcui/env/anaconda3/envs/stCAMBL/lib/R'

Read data and perform stCAMBL analysis

random_seed = 2050
stCAMBL.set_seed(random_seed)
device = 'cuda:1' if torch.cuda.is_available() else 'cpu'
root = '/data3/yfchen/stCAMBL/data/nanostring'
ids = [
    'fov1', 'fov2', 'fov3', 'fov4', 'fov5',
    'fov6', 'fov7', 'fov8', 'fov9', 'fov10',
    'fov11', 'fov12', 'fov13', 'fov14', 'fov15',
    'fov16', 'fov17', 'fov18', 'fov19', 'fov20'
]
dataset = ids[16]
adata = sc.read_h5ad(os.path.join(root, dataset, 'sampledata.h5ad'))
n_clusters = len(set(adata.obs['merge_cell_type']))
# add ground_truth
adata.obs['ground_truth'] = adata.obs['merge_cell_type']
adata.layers['count'] = adata.X
sc.pp.filter_genes(adata, min_cells=50)
sc.pp.filter_genes(adata, min_counts=10)
sc.pp.normalize_total(adata, target_sum=1e4)
sc.pp.highly_variable_genes(adata, flavor="seurat_v3", layer='count', n_top_genes=2000)
adata = adata[:, adata.var['highly_variable'] == True]

sc.pp.scale(adata)
if scipy.sparse.issparse(adata.X):
    adata.X = adata.X
from sklearn.decomposition import PCA
adata_X = PCA(n_components=200, random_state=42).fit_transform(adata.X)
adata.obsm['X_pca'] = adata_X
graph_dict = stCAMBL.graph_construction(adata, 12)
model = stCAMBL.stCAMBL(dataset, adata.obsm['X_pca'], graph_dict, device=device , rec_w=6, gcn_w=6, self_w=4, hsl_w=4, csl_w=1)
model.train_model(epochs=200)
stCAMBL_feat, defeat, _, _, _ = model.process()
adata.obsm['emb'] = stCAMBL_feat
radius = 50
tool = 'mclust'
from stCAMBL.clust_func import clustering
clustering(adata, n_clusters, radius=radius, method=tool, refinement=True)
sub_adata = adata[~pd.isnull(adata.obs['merge_cell_type'])]
ARI = metrics.adjusted_rand_score(sub_adata.obs['merge_cell_type'], sub_adata.obs['domain'])
adata.uns['ARI'] = ARI
print('Dataset:', dataset)
print('ARI:', ARI)

100%|██████████| 200/200 [00:46<00:00,  4.28it/s]

fitting ...
  |======================================================================| 100%
Dataset: fov17
ARI: 0.5481155777764808

Draw cluster map and UMAP

adata.obsm['spatial'][:, 1] = -1*adata.obsm['spatial'][:, 1]
plt.rcParams["figure.figsize"] = (10, 8)
plt.rcParams['figure.dpi'] = 300
plt.rcParams['savefig.dpi'] = 300
plt.rcParams['font.family'] = 'DejaVu Sans'
plot_color=["#F56867","#556B2F","#C798EE","#59BE86","#006400","#8470FF",
           "#CD69C9","#EE7621","#B22222","#FFD700","#CD5555","#DB4C6C",
           "#8B658B","#1E90FF","#AF5F3C","#CAFF70", "#F9BD3F","#DAB370",
          "#877F6C","#268785", '#82EF2D', '#B4EEB4']
ax = sc.pl.embedding(adata, basis="spatial",
                    color=['domain'],
                    s=120,
                    alpha=0.9,
                    edgecolor='black',
                    linewidth=0.3,
                    show=True,
                    palette=plot_color,
                    title='stCAMBL',)