Beyond benchmarking and towards predictive models of dataset-specific single-cell RNA-seq pipeline performance

This repository includes:

1. clusters.zip
   • .Once unzipped, the top-level directory is named clusters/.
     • Within this directory, there are 86 sub-directories, one for each of the scRNA-seq datasets used in the paper. These sub-directories are named with the corresponding EBI Single Cell Atlas IDs.
       • Each sub-directory contains CSV files containing clustering results for each of the pipelines run on that dataset.
2. *_unscaled.csv 
   • CSV files containing raw performance metrics (CH, DB, SIL, GSEA) computed on each pipeline and dataset combination
3. *CorrectedImputed.csv and gseaScaledImputed.csv
   • CSV files containing performance metrics corrected for the number of clusters and missing values imputed in the case of CH, SIL, and DB. For GSEA, only scaling and imputation was performed.
4. pipelineParams.csv and datasetFeatures.csv
   • CSV files containing parameters used for each of the pipelines run, and dataset summary statistics for each of the scRNA-seq datasets.

Files with the prefix "large_samples" correspond to data associated with the 6 scRNA-seq datasets containing >100k cells. The large_samples_clusters.zip file contains the same structure as cluster.zip, but with 6 sub-directories.