With the advent of spatial multi-omics, we can mosaic integrate such datasets with partially overlapping modalities to construct higher dimensional views of the source tissue. SpaMosaic is a spatial multi-omics mosaic integration tool that employs contrastive learning and graph neural networks to construct a modality-agnostic and batch-corrected latent space suited for analyses like spatial domain identification and imputing missing omes.
We tested our code on a server running Ubuntu 18.04.5 LTS, equipped with 4 NVIDIA A6000 GPUs. The installation process typically takes 10–15 minutes.
git clone https://github.com/JinmiaoChenLab/SpaMosaic.git
cd SpaMosaic
conda create -n SpaMosaic python=3.8.8
conda activate SpaMosaic
pip install -r requirements.txt
# install torch
pip install torch==2.1.1+cu121 -f https://download.pytorch.org/whl/torch_stable.html
# install torch_geometrics
pip install torch_geometric==2.4.0 pyg_lib==0.3.1+pt21cu121 torch_scatter==2.1.2+pt21cu121 torch_sparse==0.6.18+pt21cu121 torch_cluster==1.6.3+pt21cu121 torch_spline_conv==1.2.2+pt21cu121 -f https://data.pyg.org/whl/torch-2.1.1+cu121.html
python setup.py install
R package mclust is needed to perform clustering and make sure it installed in a R environment.
We provided detailed tutorials on applying SpaMosaic to various integration or imputation tasks. Please refer to https://spamosaic.readthedocs.io/en/latest/.
Source of public datasets:
- Mouse embryonic brain dataset:
three slices - Mouse postnatal brain dataset (rna+atac): {
slice 1, 2}, {slice 3} - Mouse postnatal brain dataset (rna+h3k4me3): {
slice 1, 2}, {slice 3} - Mouse postnatal brain dataset (rna+h3k27me3): {
slice 1,2}, {slice 3} - Mouse postnatal brain dataset (rna+h3k27ac):
three slices - Mouse embryo: {
slice 1}, {slice 2,3,4} - Five-modal mouse brain dataset (rna+atac+histone):
four slices
We have compiled the simulation, in-house, and public datasets into h5ad files. Please refer to zenodo.
To reproduce SpaMosaic's results, please visit reproduce folder.
To reproduce compared methods' results, including CLUE, Cobolt, scMoMaT, StabMap, MIDAS, TotalVI, MultiVI, Babel, please visit https://github.com/XiHuYan/Spamosaic-notebooks.