Snakemake workflow for subsampling and repeat clustering

A snakemake pipeline to scatter and gather Seurat@ident by subsampling the cells and repeat for multiple times. This is useful for evaluating the cluster stability using different parameters.

For now, three parameters are tested.

• number of PCs (principle components) used for RunPCA (npcs) and FindNeighbors (dims).
• k.param for FindNeighbors
• resolution for FindClusters

on odyssey cluster(SLURM):

ssh odyssey

## start a screen session
screen

git clone https://github.com/crazyhottommy/pyflow_seurat_parameter

conda install -c conda-forge mamba
mamba create -c conda-forge -c bioconda -n snakemake snakemake

source activate snakemake

# R3.5.1, make sure you load R after source activate conda environment
module load R

#hdf5, seurat needs a more recent hdf5 to be able to install
module load hdf5

R
>install.package("Seurat")

If you do not want to install Seurat yourself, a singularity container image is available if you evoke snakemake (This is for Seurat V3 only):

snakemake --use-singularity

copy your seurat object into the pyflow_seuratv3_parameter folder and

open the config.ymal file to edit some configurations.

# dry run
snakemake -np 

# if on bioinfo1 or bioinfo2 (there are 64 cores avaiable on each node)
snakemake -j 40

# if submitting job to queue (for slurm only)

./pyflow-scBoot.sh

The rds output from this workflow can be directly used by the scclusteval R package for downstream analysis.