MultimetricST Framework

Multi-Metrics Framework for Spatial Transcriptomics clustering evaluation.

This repository contain metrics code from the article "Multi-Perspective Evaluation of Spatial Transcriptomics Clustering Methods"

install requirements for python==3.10.0

pip install -r requirements.txt

install MultimetricST package

pip install git+https://github.com/InfOmics/MultimetricST.git

test_mutlimetricST.py contains the MultimetricST Test pipeline on randomly generated cluster labels.

Given the predicted label and an optional.

multimetricST/run_evaluator.py
Evaluate clustering performance given an existing AnnData object and an optional raw dataset loading.
Evaluate clustering performance given raw expression and spatial matrices.

Spatial_Clustering_Methods folder contains the MultimetricST Framework for all 9 methods and the dashboard vissualizzation

Create an environment if necessary

conda create -n multimetricst python==3.10.0 r-base==4.3.1 -y

conda activate multimetricst

install requirements

pip install -r requirements2.txt

Run clustering and evaluation in folder Spatial_Clustering_Methods Clustering_Tutorial.py downloads the varous methods repos, run each method and save results to clustering_results.csv: cd Spatial_Clustering_Methods

python Clustering_Tutorial.py 

Visualize dashboard run ipynb script: runDashboard.ipynb

Note: To run the SpaceFlow method from the downloaded repo, the code In Spatial_Clustering_Methods/SpaceFlow/SpaceFlow/SpaceFlow.py line 132 need to use defualt flavour. sc.pp.highly_variable_genes(adata, n_top_genes=n_top_genes, flavor='cell_ranger', subset=True) -> sc.pp.highly_variable_genes(adata, n_top_genes=n_top_genes, subset=True)

Spatial_Clustering_Methods/SEDR/SEDR/clustering_func.py line 52

Data Availability

The spatial transcriptomics datasets are available at: https://zenodo.org/records/17167458

Download the DLPFC 151673 data used in Clustering_Tutorial.py : wget https://zenodo.org/records/17167458/files/Data.zip