STEM is a tool for building single-cell level spatial transcriptomic landscapes using SC data with ST data. STEM extracts the spatial information from the gene expressions and eliminates the domain gap between spatial transcriptomics and single-cell RNA-seq data.
conda install pytorch torchvision torchaudio pytorch-cuda=11.7 -c pytorch -c nvidia ## more info: https://pytorch.org/get-started/locally/
conda install -c conda-forge scanpy python-igraph leidenalg
conda install seabornSince STEM is a light model, you can use the code from the STEM folder directly. You can also install STEM from PYPI:pip install scSTEM.
To verify the installation, run python test.py
Then we will demonstrate the workflow of generating the results shown in our figure 2.
%pylab inline
import os
os.environ["CUDA_VISIBLE_DEVICES"] = "3"
import scanpy as sc
import pandas as pd
import torch
import scipy
import time
from STEM.model import *
from STEM.utils import *First, we load the simulated ST data as ST data and the raw SeqFISH data as SC data. Then we normalized and log-scaled these data.
scdata = pd.read_csv('./data/mousedata_2020/E1z2/simu_sc_counts.csv',index_col=0)
scdata = scdata.T
stdata = pd.read_csv('data/mousedata_2020/E1z2/simu_st_counts.csv',index_col=0)
stdata = stdata.T
stgtcelltype = pd.read_csv('./data/mousedata_2020/E1z2/simu_st_celltype.csv',index_col=0)
spcoor = pd.read_csv('./data/mousedata_2020/E1z2/simu_st_metadata.csv',index_col=0)
scmetadata = pd.read_csv('./data/mousedata_2020/E1z2/metadata.csv',index_col=0)
adata = sc.AnnData(scdata,obs=scmetadata)
sc.pp.normalize_total(adata)
sc.pp.log1p(adata)
scdata = pd.DataFrame(adata.X,index=adata.obs_names,columns=adata.var_names)
stadata = sc.AnnData(stdata)
sc.pp.normalize_total(stadata)
sc.pp.log1p(stadata)
stdata = pd.DataFrame(stadata.X,index=stadata.obs_names,columns=stadata.var_names)
adata.obsm['spatial'] = scmetadata[['x_global','y_global']].values
stadata.obsm['spatial'] = spcoorNext we calculate the ratio between the median total counts value of SC and ST data as the dropout rate.
sc.pp.calculate_qc_metrics(adata,percent_top=None, log1p=False, inplace=True)
adata.obs['n_genes_by_counts'].median()
sc.pp.calculate_qc_metrics(stadata,percent_top=None, log1p=False, inplace=True)
stadata.obs['n_genes_by_counts'].median()
dp = 1- adata.obs['n_genes_by_counts'].median()/stadata.obs['n_genes_by_counts'].median()
#0.5836734693877551We first config the STEM model and then train it. Empirically we found by setting the sigma as half of the ST spot adjacent distance, STEM achieves the best performance.
class setting( object ):
pass
seed_all(2022)
opt= setting()
setattr(opt, 'device', 'cuda:0') # device
setattr(opt, 'outf', 'log/test') # folder to save log files
setattr(opt, 'n_genes', 351) # number of genes for the input
setattr(opt, 'no_bn', False) # duplicated
setattr(opt, 'lr', 0.002) # learning rate
setattr(opt, 'sigma', 3) # the spatial variance parameter in the Gaussian function
setattr(opt, 'alpha', 0.8) # MMD loss weight default:0.8
setattr(opt, 'verbose', True) # verbose
setattr(opt, 'mmdbatch', 1000) # batch for MMD loss
setattr(opt, 'dp', dp) # dropout rate for ST data
testmodel = SOmodel(opt)
testmodel.togpu()
loss_curve = testmodel.train_wholedata(400,torch.tensor(scdata.values).float(),torch.tensor(stdata.values).float(),torch.tensor(spcoor.values).float())The loss curve will be like this:
We first get the embeddings and build the mapping matrix We first get the embeddings and build the mapping matrix
testmodel.modeleval()
scembedding = testmodel.netE(torch.tensor(scdata.values,dtype=torch.float32).cuda())
stembedding = testmodel.netE(torch.tensor(stdata.values,dtype=torch.float32).cuda())
netst2sc = F.softmax(stembedding.mm(scembedding.t()),dim=1).detach().cpu().numpy()
netsc2st = F.softmax(scembedding.mm(stembedding.t()),dim=1).detach().cpu().numpy()The matrix netst2sc and netsc2st are ST-SC and SC-ST mapping matrices, respectively. In the ST-SC mapping matrix, the probability of one spot to other cells summarizes to 1. In the SC-ST mapping matrix, the probability of one cell to other spots summarizes to 1.
Then we can get the spatial coordinate for every single cell.
adata.obsm['spatialDA'] = all_coord(pd.DataFrame(netsc2st,index=adata.obs_names,columns=stadata.obs_names),spcoor)Compared with other methods, STEM is the only method that preserves the original topology structure of all single cells.
More demos can be found in the Demo folder. Codes for repruducing the results are in the SourceforFigure folder.
The processed data and trained models used for reproducing the results are deposited in Figshare.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent