Compositional Covariance Shrinkage and Regularised Partial Correlations

This is the code repository for the paper "Suzanne Jin, Cedric Notredame, & Ionas Erb. (2022) Compositional Covariance Shrinkage and Regularised Partial Correlations".

In this repository you can find:

• The data we used for the benchmark. It is a single-cell gene expression dataset generated from a poupulation of asynchronized mouse stem cells (Riba et al., 2022).

  • Inside the folder data/genes244 you can find the subset of 244 nonzero genes across all cells. This was used to simulate synthetic logistic normal data.
  • Inside the folder data/genes770 you can find the subset of 770 genes for which 3986 cells out of the total of 5637 cells have no zeros.
  • Both folders contain the corresponding count matrix count.csv.gz, the gene names features.csv and the cell barcodes barcodes.csv. The count matrix is organized in such a way so that rows are cells and columns are genes.

• The results produced from the synthetic data in results/genes244

• A R script bin/benchmark-simulation.R that we used to produce the results from the synthetic data.

• A Nextflow pipeline that we used to produce the main results with and without imputation. This pipeline has many components:

  • bShrink.nf. This is the main script.
  • modules folder contains the Nextflow modules that calls the R scripts stored in bin.
  • nextflow.config. This is the main config file for Nextflow.
  • conf folder contains the rest of config files.

To run the Nextflow pipeline, you should:

1. Install Nextflow. Please check https://www.nextflow.io/
2. Install Singularity. This will allow you to run the scripts inside a container with all the preinstalled packages.
3. Run Nextflow pipeline:

nextflow run -profile singularity bShrink.nf

Then, you can find the results in results/genes770. You can also reproduce the figures from the paper with the jupyter notebooks.

References

Jin, S., Notredame, C., & Erb, I. (2022). Compositional Covariance Shrinkage and Regularised Partial Correlations. arXiv preprint arXiv:2212.00496.

Riba, A., Oravecz, A., Durik, M., Jiménez, S., Alunni, V., Cerciat, M., ... & Molina, N. (2022). Cell cycle gene regulation dynamics revealed by RNA velocity and deep-learning. Nature Communications, 13(1), 2865.