create null sequences with gkmSVM for mouse genome mm10

Question

[fusion.slope]

Hello,

does anybody of you has ever tried to generate null model of DNA sequences in mouse using the gkmSVM package? It works perfectly for human but for mouse it does not. I am wondering if any of you have ever used this package for mice and had the same problem.

The error problem when I use the genNullseqs function is:

library(gkmSVM)
library(BSgenome.Mmusculus.UCSC.mm10)
library(BSgenome.Mmusculus.UCSC.mm10.masked)
library(IRanges)

genome=BSgenome.Mmusculus.UCSC.mm10.masked

fileBedBreaks="Rep1.intersec.Rep2.cov.major2.sort.uniq.bed"
fileFastaPos="Rep1.intersec.Rep2.cov.major2.all.sort.uniq.bed.pos.fa"
fileBedNeg="Rep1.intersec.Rep2.cov.major2.all.sort.uniq.Random.gkmSVM.bed"
fileFastaNeg="Rep1.intersec.Rep2.cov.major2.all.sort.uniq.Random.gkmSVM.fa"
genNullSeqs(inputBedFN=fileBedBreaks,nMaxTrials=5,xfold=2,genome=genome,
            outputPosFastaFN=fileFastaPos,outputBedFN=fileBedNeg,outputNegFastaFN=fileFastaNeg)

Error in normalizeDoubleBracketSubscript(i, x, exact = exact) : subscript "TRF" matches no name

TRF is related with the tandem repeats because it means Tandem Repeats Finder. But what does it means the problem??

Thanks in advance for any reply

Answer 1

[sarahmcclymont]

I hit this same problem and figured out that it's because the BSgenome.Mmusculus.UCSC.mm10.masked on Bioconductor (as of May 2020) only has two masks included (AGAPS and AMB), whereas previous genomes included the TRF mask as well, which is what gkmSVM is looking for.

Thankfully, on GitHub, Zhang (Frank) Cheng has added the TRF mask to the mm10 genome at https://github.com/biomystery/BSgenome.Mmusculus.UCSC.mm10.masked

I had to uninstall the Bioconductor version of mm10 and reinstalled using remotes::install_github("biomystery/BSgenome.Mmusculus.UCSC.mm10.masked")

[fusion.slope]

awesome! now null sequences model also for the mouse genome!