Hi:

I have set of genomic interval in GRanges objects. However I intend to filter them conditionally, while export them as bed format files. I have implemented function that possibly accomplish this task, but I have an error. I  have used dplyr packages for facilitating output more efficiently. I simulated same data set as data.frame and it worked perfectly. How can I fix this issue? Any quick solution ?

mini example:

grs <- GRangesList(
  foo = GRanges( seqnames=Rle("chr1", 4),ranges=IRanges(c(3,33,54,91), c(26,42,71,107)),
                  rangeName=c("a1", "a2", "a3", "a4"), score=c(22, 6,13, 7)),
  bar = GRanges(seqnames=Rle("chr1", 6),ranges=IRanges(c(5,12,16,21,37,78), c(9,14,19,29,45,84)),
                 rangeName=c("b1", "b2", "b3", "b4", "b5", "b6"), score=c(3, 5, 3, 9, 4, 3)),
  cat = GRanges(seqnames=Rle("chr1", 7),ranges=IRanges(c(1,8,18,35,42,59,81), c(6,13,27,40,46,63,114)),
                 rangeName=c("c1", "c2", "c3", "c4","c5","c6","c7"), score= c(2.1, 3, 5.1, 3.5, 7, 2, 10))
)

helper function :

.cast.Pvl<- function(x, pvalueBase = 1L, ...) {
  if(is.null(x$pvalue)){
    x$p.value <- 10^(score(x)/(- pvalueBase))
    colnames(mcols(x))[3] <- "p.value"
  } else {
    x
  }
  return(x)
}

I intend to implement this sketch function :

mylist2 <- lapply(seq_along(grs), function(i) {
  if(is.null(i$p.value)) {
    i <- .cast.Pvl(i, 1L)
  } else {
    splitter <- function(i, tau.w) {
      require(dplyr)
      DF <- grs[[i]]
      DF <- as(DF, "data.frame")
      DF %>%
        filter(p.value >= tau.w) %>%
        write.csv(., sprintf("dropped.%s.csv", i), row.names = FALSE)
      total.ERs <- filter(DF, p.value <= tau.w)
      return(total.ERs)
    }
  }
  res <- splitter(i, 1.0E-04)
})

I have an error like this:

Error in i$p.value : $ operator is invalid for atomic vectors

can't figure out what is possible reason. Can anyone tell me what's going on? How can I fix this bug ? How can I facilitate the output of myList as I expected ? Thanks a lot

I don't full understand what you're trying to achieve, but you're getting this error as a result of the seq_along() function.  If you run this in isolation you'll see it gives you an sequential integer vector of the same length as grs e.g.

> test <- seq_along(grs)
> test
[1] 1 2 3

Your code then applies the function to each element of that vector.  Since each element is a single integer it doesn't have a p.value element and indeed $ isn't valid on vectors e.g.

> test[1]$p.value
Error in test[1]$p.value : $ operator is invalid for atomic vectors

If you want to perform your function on each member of grs you can simply leave out the seq_along() part.  Something like

mylist2 <- lapply( grs, function(i) {
  if(is.null(i$p.value)) {
    i <- .cast.Pvl(i, 1L)
  }
})

Use the debugging features of R and investigate the type of i. It's probably not what you want. Also, I would recommend against writing your own BED exporter, because it's easy to make mistakes and more generally makes your code more complex. In this case, mistakes include not subtracting 1 from the start values (to make 0-based ranges), and using comma instead of tab as the separator.