A description of the concepts and example runs with dummy data is available in the pdf and the description folder (corresponding knitr Rnw).

There is no installation required, the R scripts that can be run on their own but do have some dependencies. Compile the C code in the scripts directory with the following command:

R CMD SHLIB scoringlite.c

This will generate the dynamic library to make co-clustering matrices from hard assignments vectors.

R packages: BiocGenerics S4Vectors IRanges GenomeInfoDb GenomicRanges

run Step1.submitALL.R with:

Rscript Step1.submitALL.R

This pipeline goes through the PCAWG sample IDs and submit one job per sample on a slurm-based cluster to run CICC.
The job will run runCICC.R, which loads the required data using utility functions from loadData.R, where paths are encoded, and then runs CICC and saves the output in a consensus format using functions from loadData.R.

run Step2.plotClusterCCF.R with:

Rscript Step2.plotClusterCCF.R

This is useful for visualisation of the results. It plots histograms of CCF and the cluster positions for each method and for the consensus.

1. writeResultsMA will write a vector of mutation assignments, i.e. integer values corresponding to consensus cluster IDs
2. writeResultsClusterCCF will write a consensus subclonal architecture: a dataframe with three columns ("cluster", "n_ssms","proportion") and as many rows as identified (sub)clones.

• cluster is the cluster ID
• n_ssms is the number of SNVs assigned to that cluster
• proportion is the fraction of cancer cells sharing these SNVs

3. writeResultsClusterCCF2 will write a file very similar to writeResultsClusterCCF but with fraction of the total cells instead of fraction of cancer cells.