Question regarding extractSignatures command

[ghost]

Hi @PoisonAlien,

maftools package is brilliant. Have a small question about extractSignatures though. Is there any particular reason for absence of nrun argument with this function? It seems to be already in use with estimateSignatures.

Since you are using nmf underneath, only 1 run is performed by default for the chosen rank. Alternative extract_signatures function from MutationalPatterns package already uses 200 runs by default.

[PoisonAlien]

Hello,
Thanks for using maftools and I am glad that you find it useful.
In my experience nrun has an insignificant influence on the overall results. The critical step is estimateSignatures and as long as you have this approximation, it should be fine for the main extractSignatures. nrun indeed helps when the data has lots of observation (as in DNAm or gene expression), since here there are only 96 channels, it should not affect the results much.

Also, you are welcome to test this by adding nrun = 100 to this line of the extractSignatures script :)

Hope this helps.

[ghost]

Makes sense. Thank you.

[ShixiangWang]

There is a function sig_extract() in sigminer which is similar to extractSignatures() in maftools, you can directly use it.

[doc, source code]

[ghost]

Thank you. I was actually checking out sigminer last night and might just use this function. For now though, I have added the nrun argument to extractSignatures with 50 runs by default, and an argument to get the absolute exposures as well. By default, extractSignatures only returns the adjusted exposures.