SLOPER is a computational framework for modeling spatial transcriptomics data with inhomogeneous Poisson point processes (IPPPs) and estimating the spatial gradient of gene expression using truncated score matching.
We provide a complete interactive walkthrough in the notebook using the DLPFC Visium dataset (#151673):
See demo.ipynb to learn
(i) how to train SLOPER to estimate the spatial gradient, and
(ii) how to run annealed Langevin dynamics to generate enhanced features.
To run demo.ipynb and the core SLOPER pipeline, you will need the following key libraries (versions used in our examples are listed below):
- numpy: 2.1.0
- pandas: 2.2.3
- scanpy: 1.11.1
- torch: 2.5.1+cu124
- shapely: 2.1.1
- scikit-learn: 1.5.2
- matplotlib: 3.10.1
- scipy: 1.15.2