Signatures Contributing to Driver Genes

[mkinnaman]

I keep throwing the same error every time I try to plot the drivers vs signatures plot:

matprob <- matrix(nrow=length(drivers),ncol=length(SBS_OSCE1),dimnames=list(drivers, SBS_OSCE1))
sig.cols <- paste0(rownames(SBS_OSCE1_sigs),".prob")#grep("prob",colnames(vcf.cod))
for(i in 1:nrow(matprob)){

• g <- rownames(matprob)[i]
• ind <- which(vcf.cod$gene_name==g)
• matprob[i,] <- apply(vcf.cod[ind,sig.cols],2,sum,na.rm=T)
• }
  Error in matprob[i, ] <- apply(vcf.cod[ind, sig.cols], 2, sum, na.rm = T) :
  number of items to replace is not a multiple of replacement length

Any thoughts?

[mkinnaman]

In addition - with my last couple of runs - have been throwing errors during the denovo signature command: Is this due to small sample size?

Error: NMF::nmf - invalid argument 'rank': must be a single numeric value
In addition: Warning messages:
1: In (function (...) :
NAs were produced due to errors in some of the runs:
-#4[r=5] -> elements of 'k' must be between 1 and 4 [in call to 'cutree']
-#5[r=6] -> elements of 'k' must be between 1 and 4 [in call to 'cutree']
-#6[r=7] -> elements of 'k' must be between 1 and 4 [in call to 'cutree']
-#7[r=8] -> elements of 'k' must be between 1 and 4 [in call to 'cutree']
-#8[r=9] -> elements of 'k' must be between 1 and 4 [in call to 'cutree']
-#9[r=10] -> elements of 'k' must be between 1 and 4 [in call to 'cutree']
2: Removed 25 rows containing missing values (geom_path).
3: Removed 62 rows containing missing values (geom_point).
4: In max(abs(diff(z))) : no non-missing arguments to max; returning -Inf

[FunGeST]

Hi,

Thanks for getting in touch, and I apologise for not getting back to you sooner.

Regarding your first issue, I think it may just be caused by SBS_OSCE1 and rownames(SBS_OSCE1_sigs) being different lengths, although without your data I can't be 100% sure.

I'm less sure about your second issue - out of interest how small is your sample size? NMF extractions should work for smaller sample sizes. To get around this error you could try specifying the num_of_sigs = argument in the NMF_Extraction() function as an integer.

If the file NMF_Rank_Estimates.pdf has been plotted in your results directory, the value in the x-axis for the first minima of the cophenetic plot is a good estimate for the optimal number of signatures in your input data. In this example, Palimpsest would choose 5 as the optimal number of signatures in the data.

Please let us know if this doesn't work and I'll try to find another solution!

Best wishes,
Benedict