Potentially mis-specified selection in `emcncf()`

[jonasfreimuth]

Hello everyone,

first of all, thanks for providing this handy package!

When troubleshooting my own issues with purity being reported as NA, I noticed that when checking for a pure and diploid tumor (

facets/R/facets-emcncf.R

Line 58 in f3c93ee

if(all(seg$tcn[seg$chrom<nX]==2 & seg$lcn[seg$chrom<nX]%in%c(1, NA))|max(mafR.clust[seg$chrom<nX & seg$nhet>min.nhet], na.rm = TRUE) < 0.05){

), the check whether any of the cluster mafR is below the threshold of 0.05 has a misspecified selection vector. The mafR.clus object is a vector of length(unique(seg$segclust)), but the selection vector has nrow(seg) elements. In the worst case, this could lead to a sample being reported as diploid and no tumor purity being reported when that is not acutally the case, but seg$chrom < nX & seg$nhet > min.nhet happened to be FALSE for all elements where the mafR.clust vector was not NA (i.e., when all the segments, whose id/position corresponded to a segclust ID/position with a non-NA mafR.clust value, happend to be below the nhet threshold). I don't know how likely this is, but it can't be ruled out that some of the open issues reporting Warning message: In max(mat$chrom) : no non-missing arguments to max; returning -Inf (which is the result of max not getting any non-NA input), namely #39, #46, and #86, may have encountered this issue when it can't be ruled out that it's strictly due to not enough segments passing the nhet threshold.

In any case, subsetting by a vector that is almost always too long is something that is hard to spot when reading the code, so as far as I can tell this should be fixed. I'd propose just replacing it with seg$mafR.clust[seg$chrom < nX & seg$nhet > min.nhet], and getting rid of the the mafR.clust object entirely, since AFAICT, the mafR.clust column in the seg dataframe is already uniform for each cluster.