Same mutation in two different signatures

[rahulk87]

Hi,

I have been running Palimpsest on mutations in whole genome and same mutations but falling in specific regions of the genome. I observed that in these two different runs, same mutations annotated with two different signature i.e. mutation X annotated with signature SBS5 when I use whole genome mutations and same mutation X annotated with signature SBS84 when I use a subset of mutations in a specific genomic region.

Is there any plausible reason for this?

Thanks!!

[jayendrashinde91]

Hi @rahulk87
Assigning the probability of mutational signature to every somatic mutation depends on the number of mutations attributed to that mutational signature. This would vary in your two runs affecting the probabilities of associated mutational signatures. Please refer our paper for more details about the methods used in Palimpsest - https://www.nature.com/articles/s41467-017-01358-x

[rahulk87]

Thanks @jayendrashinde91 for your prompt reply, I read the methods in paper.

Actually, that's exactly my question is i.e. around 1000 mutations were attributed to signature SBS84 in sample X when I used mutations from specific genomic regions and when I used the whole genome mutations of that very sample, 0 mutations were attributed to SBS84, despite input contain those ~1000 mutations, how is it possible. This is the step before applying function "signature_origins". I used the object created by function "deconvolution_fit". Quite possible that I am missing something.

[FunGeST]

Hi,

Sorry for the late reply. The probability of a specific mutation being attributed to a given signature depends on the overall signature decomposition for the sample. So if you change the input (by selecting a subset of mutations), the proportion of signatures will change, which ultimately can modify the probability of a particular mutation being due to each process. Regarding why signature composition varies when you select only a subset of mutations, this can be explained by biological reasons (mutational processes are affected by various features including transcription or replication in a signature-dependent way) or technical reasons (e.g. the number of mutations considered is low and the signature less reliable).

Hope this helps,
Eric

[rahulk87]

Thanks Eric!!