DSSC

A model-based constrained deep learning clustering approach for spatial-resolved single-cell data

Dependencies in Python

Python 3.8.1

Pytorch 1.6.0

Scanpy 1.6.0

SKlearn 0.22.1

Numpy 1.18.1

h5py 2.9.0

munkres 1.1.4

dgl 0.8.0

All experiments of DSSC in this study are conducted on Nvidia Tesla P100 (16G) GPU.

#The input data should be in h5 format with:
(1) "X" - count matrix
(2) "Y" - true labels (if available)
(3) "Pos" - spatial coordinate
(4) "Genes" - feature names (Use to build constraints)

Dependencies in R

R 4.1.0

Seurat 4.2.0

cccd 1.5

rhdf5 2.38.1

ggplot2 3.3.6

Run DSSC

1. Build constraints (See make_links_from_Markers.R)
2. Run DSSC (See run_DSSC.sh, then run.sh)

Cite this work

Lin, X., Gao, L., Whitener, N., Ahmed, A., & Wei, Z. (2022). A model-based constrained deep learning clustering approach for spatially resolved single-cell data. Genome Research, 32(10), 1906-1917.