Hi.

We are trying to investigate the mutational clusters in various TCGA cancers, and I am wondering which R package you suggest to count them.

In fact, we want to measure the no. of defined mutations, like C>T, within specified genomic distances.

I would appreciate your kind help.

Here are a few ideas, given that you have your mutations represented in a VRanges object (VariantAnnotation package):

Defined mutations: If you are only interested in e.g. C>T substitutions, you can subset your data along the lines of

vr_filt = subset(vr, ref(vr) %in% "C" & alt(vr) %in% "T")

and continue your analysis with this subset.

Visualization: You can have a look at the plotRainfall function in the SomaticSignatures packages, and its documentation. It can give you an idea where interesting regions may be, and can serve as a starting point for an exploratory analysis.

Binning: A simple, yet effective analysis for kataegis can consist of counting the number of mutations in bins along the genome. Here, you can leverage the extensive Ranges infrastructure with Bioconductor.

## divide the humen GRCh37 genome into 1 Mbp bins
bins = tileGenome(seqinfo(BSgenome.Hsapiens.1000genomes.hs37d5), tilewidth = 1e6, cut.last.tile.in.chrom = TRUE)
## count the number of mutations per bin
counts = countOverlaps(vr_filt, bins)

Distance between neighboring mutations: You can also compute the distant between adjacent mutations, and identify kataegis events this way. You may have a look at the gaps method in GenomicRanges or the mutationDistance function in SomaticSignatures for this; however, how you define this distance will also depend to some extend of the types of variants you are looking at.

Thanks very much, Julian.
Would try it!