vcfcomp returns NAs about vcfppr

[kiran-lee]

Thank you so much for our email discussion troubleshooting my NAs.

We realised this was due to missing genotypes in the "truth" vcf file. Removing missing genotypes in the "truth" dataset beforehand produced a sensible r2 value. This was the script used:

library(vcfppR)

truth <- vcftable("x.vcf.gz", setid = T, info = F)
test <- vcftable("y.vcf.gz", setid = T, format = "DS", info = F)

N <- 22 ### number of samples is 22
truth$gt <- truth$gt[,1:N]
truth$samples <- test$samples[1:N]

remove sites with genotypes being NA in truth

w <- which(!is.na(rowSums(truth$gt)))
truth$gt <- truth$gt[w,]
truth[2:9] <- lapply(truth[2:9], function(a) a[w])
saveRDS(truth, file = "truth.rds")

names <- truth$samples[1:22]
samples <- paste0(names, collapse = ",")

res <- vcfcomp(test = "y.vcf.gz" ,
truth = "truth.rds",
stats = "r2",
names = names,
samples = samples,
bins = c(0,1))

str(res)
res

[Zilong-Li]

Hi Kiran,

I think the input region is not valid, which should be in the form of chr:start-end, or just chr.

I'll add validators for the input in next release. Thanks for the feedback.

[Zilong-Li]

Also make sure the af.tsv is valid, which should have a header line and five columns. If the af is calculated from the truth file, you can just make vcfcomp doing the job by ignoring the af option ie af=NULL