biolord's reproducibility repository

This code contains the code to reproduce the results shown in "Biological representation disentanglement of single-cell data" (2023).

Data availability

1. perturbation data (sci-Plex 3, Srivatsan et al. 2020):
   1. Anndata file used for the initial pre-processing notebook: sciplex_complete_middle_subset.h5ad (provided by Hetzel et al.).
   2. Processed anndata file used for the analysis: sciplex3_biolord.h5ad.
2. perturbation data (Perturb-seq (1-gene), Adamson et al. 2016):
   1. Data used for the initial pre-processing notebook: Downloaded from here using GEARS within the notebook.
   2. Processed anndata file used for training a biolord model analysis: adamson_biolord.h5ad and adamson_single_biolord.h5ad.
3. perturbation data (Perturb-seq (2-gene), Norman et al. 2019):
   1. Data used for the initial pre-processing notebook: Download the data from here. Move the uncompressed norman2019.tar.gz folder to ./data/perturbations/norman. It should contain the subdirectory data_pyg. Move essential_norman.pkl and go_essential_norman.csv to ./norman, So ./data/perturbations/norman should contain essential_norman.pkl, go_essential_norman.csv and norman2019/data_pyg/.
   2. Processed anndata file used for training a biolord model analysis: norman_biolord.h5ad and norman_single_biolord.h5ad.
4. fetal chromatin accessibility atlas (Domcke et al.):
   1. Data used for the initial pre-processing notebook: Domcke-2020.h5ad (provided by Cao et al.).
   2. Processed anndata file used for training a biolord model analysis: atac_biolord.h5ad.
5. spatio-temporal infection dataset (Afriat et al.):
   1. Data used for the initial pre-processing notebook: 10.5281/zenodo.7081862 (provided by Afriat et al.)
   2. Processed anndata file used for the infection analysis: adata_infected.h5ad.
   3. Processed anndata file used for the abortive classification analysis: adata_abortive.h5ad.