This repo contains the code used to run panpipes benchmarks. To allow a fair comparison, equivalent methods where run, namely scvi, harmony and PCA for no batch correction.
We also ran a panpipes only benchmark for time and resources with scalability of number of cells in the following datasets:
- Seurat Multimodal PBMC (RNA.ADT) post-QC at 10K,50K,100K,152K (Seurat v4 paper)
- TARUS dataset (RNA only) at 500K and 1M ( biorxiv preprint)
The R script was submitted using the provided bash submission script and elapsed times are printed to a log file. The time is then calculated as the difference in seconds between the start and the end of the computation.
Panpipes integration was run calling
panpipes integration make full
For the seurat benchmarks, we count the run time until the process collate_integration_outputs which is comparable with the end of the provided R script as it's the step of the pipeline that collates the output of all the integration methods run in parallel.