README.md

bmDA

The implementation of benchmarking the performance of differential abundance (DA) testing methods including 5 clustering-free methods:

1. Testing for differential abundance in mass cytometry data (Cydar)
2. Detection of differentially abundant cell subpopulations in scRNA-seq data (DAseq)
3. Quantifying the effect of experimental perturbations at single-cell resolution (MELD)
4. Differential abundance testing on single-cell data using k-nearest neighbor graphs (Milo)
5. Co-varying neighborhood analysis identifies cell populations associated with phenotypes of interest from single-cell transcriptomics (CNA),

and a clustering-based method, Louvain.

Dependencies

To run this benchmarking codes, it needs to install a list of R and Python packages. The R packages needed are:

• argparse
• SingleCellExperiment
• scran
• DAseq
• miloR
• tibble
• dplyr
• tidyverse
• igraph
• cydar
• pdist
• reshape2

The Python packages needed are

• MELD
• cna
• scanpy
• graphtools
• scikit-learn
• multianndata

Data

• Synthetic datasets are available under the data directory.
• Real dataset are available at this link.

Usage

Note: Our implementation can only be used on a cluster with Slurm job scheduler since we need to run thousands of jobs.

The benchmarking scripts are all located in the bin drectory.

bin
├── bm_parameter.sh
├── bm_runtime.sh
├── bm_syn_real.sh
└── make_bm_data.sh

To run a benchmarking job, use the following command:

bash bm_{the script}.sh

Acknowledgement

Our implementation is inspired by the repo https://github.com/MarioniLab/milo_analysis_2020.