Hello folks,

I am pondering about an issue with the follow and precede functions GenomicRanges. I have a GRanges object with CpG-Island coordinates and another large GRanges object with the coordinates of pretty much all CpGs outside islands:

rangedCpGs
GRanges object with 29467879 ranges and 0 metadata columns:
             seqnames                 ranges strand
                <Rle>              <IRanges>  <Rle>
         [1]     chr1     [3000574, 3000575]      *
         [2]     chr1     [3000575, 3000576]      *
         [3]     chr1     [3000726, 3000727]      *
         [4]     chr1     [3001395, 3001396]      *
         [5]     chr1     [3001631, 3001632]      *
         ...      ...                    ...    ...
  [29467875]     chrX [166649576, 166649577]      *
  [29467876]     chrX [166649598, 166649599]      *
  [29467877]     chrX [166649696, 166649697]      *
  [29467878]     chrX [166649712, 166649713]      *
  [29467879]     chrX [166649815, 166649816]      *
  -------
  seqinfo: 20 sequences from an unspecified genome; no seqlengths

CpGislands
GRanges object with 74986 ranges and 0 metadata columns:
          seqnames             ranges strand
             <Rle>          <IRanges>  <Rle>
      [1]    chr10 [3001252, 3001468]      *
      [2]    chr10 [3001990, 3002096]      *
      [3]    chr10 [3102618, 3102897]      *
      [4]    chr10 [3103777, 3103948]      *
      [5]    chr10 [3104651, 3105129]      *
      ...      ...                ...    ...
  [74982]     chrY [2612251, 2612589]      *
  [74983]     chrY [2612783, 2612856]      *
  [74984]     chrY [2618798, 2619227]      *
  [74985]     chrY [2619483, 2619753]      *
  [74986]     chrY [2681236, 2681710]      *
  -------
  seqinfo: 33 sequences from an unspecified genome; no seqlengths

Now I would like to assign every CpG to it's nearest island and furthermore assess, whether it precedes or follows this particular island. 

Hence I started with

IDclosestCGI <- nearest(rangedCpGs,CpGislands)

and wanted to check if

CpGprecedesCGI <- IDclosestCGI == follow(rangedCpGs,CpGislands)

with the intention that a TRUE would mean the CGI is following the CpG. However I got a vector completely TRUE.  So I did the opposite check with precede with the same result, which I didn't expect at all.

Can somebody please explain to me, why this happens?

test <-  precede(rangedCpGs,CpGislands)
test2 <-  follow(rangedCpGs,CpGislands)
head(test)
[1] 31230 31230 31230 31230 31230 31230
head(test2)
[1] 31230 31230 31230 31230 31230 31230
all(test==test2)
[1] TRUE
sum(overlapsAny(rangedCpGs,CpGislands))
[1] 0

Thanks a lot for reading and potentially your help.

Matthias

 sessionInfo()
R version 3.1.1 (2014-07-10)
Platform: x86_64-unknown-linux-gnu (64-bit)

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8
 [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8
 [7] LC_PAPER=en_US.UTF-8       LC_NAME=C
 [9] LC_ADDRESS=C               LC_TELEPHONE=C
[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C

attached base packages:
[1] stats4    parallel  stats     graphics  grDevices utils     datasets
[8] methods   base

other attached packages:
[1] GenomicRanges_1.18.4 GenomeInfoDb_1.2.5   IRanges_2.0.1
[4] S4Vectors_0.4.0      BiocGenerics_0.12.1  BiocInstaller_1.16.5

loaded via a namespace (and not attached):
[1] tools_3.1.1   XVector_0.6.0

Most likely need an ignore.strand=TRUE on those calls. This is a really confusing consequence of strand-dependence. Yes, those features are unstranded, but to GRanges, that apparently means "either direction" instead of "left-right". It's tough to imagine cases where one would want it to mean "either direction", but that's how it works, I guess.