spatialsurf 2.0

A self-supervised deep learning method for reference-free deconvolution.

SURF

A self-supervised deep learning method for reference-free deconvolution. The overall approach is detailed in the official paper out in xxx.

Data input

df_expr: (dataframe), column names: gene names, shape: (n_spots, n_genes). The gene expression of ST data.
df_pos: (dataframe), column names: ‘x’, ‘y’, shape: (n_spots, 2). The position data of ST data.
barcodes: (list), len: n_spots. The barcodes of ST data.

Installation

We have tested the installation process under ubuntu 22.04, R 3.6.3, and torch 1.11+cuda 11.2.

1. Install R environment (https://cran.r-project.org/)
2. Create the virtual environment

conda create -n SURF python=3.9   
conda activate SURF   

3. Install Pytorch (https://pytorch.org/), please choose the suitable torch version according to your cuda version.

pip install torch==1.11.0+cu113 torchvision==0.12.0+cu113 torchaudio==0.11.0 --extra-index-url https://download.pytorch.org/whl/cu113 

Note: The installation command shown above is suitable for our cuda version and is provided as an example only. Please refer to the instructions at [https://pytorch.org/get-started/previous-versions/] to find the installation command appropriate for your cuda version.

4. Install SURF

pip install spatialsurf

Tutorials

https://github.com/lllsssyyyy/SURF/tree/main/tutorials