SEDR Analyses

Here is the analysis code for SEDR project.

We tested SEDR on DLPFC dataset (12 slices) and compared it with 5 state-of-the art methods:

To run analyses code properly, we recommend you to organize working folder as shown below and download the scripts into the folder.

Benchmarking

data folder structure

SEDR
├── data
│   └── DLPFC
│        └── 151507
│              ├── filtered_feature_bc_matrix.h5
│              ├── metadata.tsv 
│              └── spatial
│                     ├── scalefactors_json.json  
│                     ├── tissue_positions_list.csv  
│                     ├── full_image.tif  
│                     ├── tissue_hires_image.png  
│                     └── tissue_lowres_image.png  
├── output      
│      └── DLPFC          
│            └── 151507 
└── SEDR_analyses
    ├── DLPFC_Seurat.R  
    └── ...

Download data

DLPFC data can be downloaded from SpatialLIBD. Extract and put data within data/DLPFC folder.
Please notice that the scale_factors_json.json and tissue_positions_list.csv can be found in 10X folder in SpatialLIBD.
Besides, the metadata.tsv we used in SEDR is consistant with BayesSpace.
For convenient, we have put three files within data folder here. You need to move the data folder to where we recommend.

Run SEDR

Follow the instructions in SEDR to run SEDR for 12 slices.
Move the results to output/DLPFC/sample.name/SEDR for further comparison.

Run state-of-the-art methods

Rscript DLPFC_Seurat.R sample n_clusters
python DLPFC_stLearn.py sample
python DLPFC_SpaGCN.py sample n_clusters
Rscript DLPFC_BayesSpace.R sample n_clusters
Rscript DLPFC_Giotto.R sample n_clusters

Table of n_clsuters:

Sample_ID	n_clusters
151507	7
151508	7
151509	7
151510	7
151669	5
151670	5
151671	5
151672	5
151673	7
151674	7
151675	7
151676	7

Compare SEDR and other methods

Rscript DLPFC_comp.R sample

Summary of 12 slices

Rscript DLPFC.ARI_boxplot.R

Stero-seq data

Stero-seq data in SEDR project is included in data folder.

josschavezf / SEDR_analyses