MASI

MASI: marker-assisted standardization and integration for single-cell transcriptomics data

Manuscript: Fast model-free standardization and integration of single-cell transcriptomics data

1. Brief description

MASI utilizes robust marker idenfication to identify marker genes from reference data and transfers cell-type labels to target data through MACA.

2. Install requirement packages

pip install scanpy cosg rpy2 
pip install sccaf ##optional
pip install fa2 ##install if doing integrative lineage analysis

##install Seurat and RobustRankAggreg separately in R
install.packages('Seurat')
install.packages('RobustRankAggreg')

##Installment of PyTorch (Optional)
##We noticed adding BatchNorm1D as data transformation can further remove batch effects 
##but may sacrifice discriminative power for cell-type identification.

##Then git clone MASI to the working directory for use.

3. Usage

import MASI as masi

##identify cell-type markers from reference data
##source is Anndata of reference data that has cell-type labels named 'cell_type'
cell_markers = masi.marker_identification(source_data=source)

##step 1 transform gene expression matrix to cell-type score matrix
##ad is combined expression data in Anndata format, and cell_markers is dict for cell-type markers 
##scores can further be used for visualization and other downstream analyses
scores, labels = masi.gene2cell(ad=ad,cell_markers=cell_markers,use_weight=True)
##step 2 clustering and parallel annotation
annotation= masi.parallel(scores=scores,labels=labels,batch_size=50000)

4. Reproduce results in manuscript

Please see tutorials at https://github.com/hayatlab/MASI/tree/main/tutorial

Processed data can be found at https://figshare.com/articles/dataset/Fast_model-free_integration_and_transfer_learning_via_MASI_for_single-cell_expression_data/18866264.

5. Citation

Xu et al. "MASI enables fast model-free standardization and integration of single-cell transcriptomics data". Communications Biology

6. Integration of multimodal single-cell and spatial data

A major update of MASI is availale at https://github.com/hayatlab/MASIv2

MASIv2 integrates large-scale multimodal single-cell and spatial data with few computational resources. bioRxiv