This is a Github repository to accompany the paper:

Risso D, Purvis L, Fletcher R, Das D, Ngai J, Dudoit S, Purdom E (2018) "clusterExperiment and RSEC: A Bioconductor package and framework for clustering of single-cell and other large gene expression datasets" PLoS Comput Biol. 2018 Sep 4;14(9):e1006378 http://dx.plos.org/10.1371/journal.pcbi.1006378

How to Reproduce the analysis

From scratch

The provided make file will allow you to recreate the analysis in the paper from scratch. This code includes downloading the data from GEO.

You should first have R installed and the necessary packages installed: (BiocInstaller,clusterExperiment,benchmarkme,and scone) ).

Then type the following commands:

make OEAnalysis.Rout

Note that this is quite computationally intensive, and runs on parallel cores (by default the value defined by the environment variable SLURM_CPUS_PER_TASK or 6 if such a value is missing). So it should only be run from scratch in a setting appropriate to this.

Using the provided intermediate results (quicker)

To have quicker access to the data and results, we also provide (via git lfs) .rda files that are the result of the output of downloading the data, and the RSEC command (see https://git-lfs.github.com/ for instructions on git lfs and how to download these files). They are saved under the directory dataOutput_submitted.

To be able to run the above make command, and not have to rerun the computationally intensive commands, you must make the following changes:

1. Move (or copy) the directory dataOutput_submitted into dataOutput

2. Edit the file OEAnalysis.R so that the line

   runClus<-FALSE
   

   now reads as

   runClus<-TRUE
   

After these changes you should be able to run the above make command,

make OEAnalysis.Rout

With these settings, the scripts should be able to run on a laptop relatively quickly.