Hello everybody,

I am using SomaticSignatures only recently. The signature estimation with the package works well. I want to compare these signatures with the one from Alexandrov work. I asked a question on bioconductor (https://support.bioconductor.org/p/61913/), but it is probably not very visible.

When I use SomaticSignatures on my dataset, I see a difference between the signatures determined by the methods of the package and the signatures from Alexandrov: the sum of each Alexandrov signature is 1 but the sum of each determined signature on my data is between 18 and 22 (sum(as.numeric(sigs_nmf@signatures[,1]))). 

Is it normal to get sum beyond 1? Should I normalize the vectors by their sum to get a "standard signature"?

Thank you for your help,

Jane

For the comparison you have in mind, you can normalize by the sum of each signature with

w = signatures(sigs_nmf)
w_norm = t(t(w) / colSums(w))

The same is done within the plotting functions of the package, e.g. if you call plotSignatures with normalize = TRUE.

Generally speaking, there is no requirement in the decomposition of the mutational spectrum that the sum for a signature or motif have to be 1. If it is (as in the reported signatures), it is the result of normalizing the matrix.

Thank you for the clarification Julian