This repository contains an example implementation in R using Seurat of the framework outlined in:
Patterson-Cross, R.B., Levin, A.J. & Menon, V. Guided selection of single cell clustering parameters through sub-sampling cluster robustness metrics. (2020).
To use the code and examples in the repository, first clone the repository to your computer:
git clone https://github.com/rbpatt2019/cluster.stability.gitThe code is implemeted in R, and the dependencies are pinned in the renv.lock file. To install dependencies, open an R terminal, then proceed as follows:
install.packages("renv") # Not necessary if already installed
renv::init()This will create a local project and install the dependencies there, rather than into your global R installation.
Two example scripts are included in this repo. The first runs through the main analysis framework and covers the key steps, including iterative, sub-sampled clustering, calculating the co-clustering frequency matrix, determining the silhouette scores, and creating the silhouette distribution plots. The second covers several additionaly visualisations that we used in the paper and find useful for understanding the patterns and clusters within your data. Both these files can be run as stand-alone scripts, as so:
Rscript examples/1_seurat_pipeline.R
Rscript examples/2_seurat_further_visuaisations.RIf used this way, they are meant to be run sequentially. Alternatively, they can be opened with any modern IDE or editor for interactive execution.
Additionally, the interested user may directly source the scripts in the R directory to create analyses that suit individual needs.
An example data set is included with this repo. It contains ~500 human PBMCs sequenced with Smart-Seq from Ding, et al. Nature Biotechnology, 2020 and is one the datasets detailed in our paper. The RDS included in the data directory has already been normalised and had PCAs calculated using Seurat's SCTransform and RunPCA functions. If you are using your own data, be sure to normalise and calculate PCAs first.