R+Gviz: Fake AlignmentsTrack() to visualize Enhancer-Promoter connections

Question

[Matthias Zepper]

Hello Biostars-Community,

I would need your kind suggestions regarding an issue with the Gviz package from Bioconductor. I am attempting plot Enhancer-Promoter connections and tried to (mis)use the AlignmentsTrack() function and the "sashimi" plots display type for it.

Typically, alignment tracks are generated from BAM files and if I do that, I can nicely display the sashimi plots as shown in the manual. However I would like to visualize enhancer-promoter connections and thus attempted to create a fake alignment file from simulated reads spanning the coordinates of the enhancers and promoters and connecting them with introns.

Unfortunately for this fake file, I can't get the sashimi plots to work. It will display the fake reads in the "pileup" mode, but when I choose type="sashimi", the track remains blank and I get a strange error. I suppose the sashimi plot depends on additional columns like the mapping quality or such, that would automatically be read from a BAM file, but which I didn't provide with the simulated data. Any suggestions?

Thanks a lot Matthias

Minimal working (not working) example:

    library("Gviz")
    library("GenomicRanges")


    eppairs <-
      structure(
        list(
          ID_trscpt = c("Ppp3cb.NM_001310426", "Ppp3cb.NM_001310427"),
          ID_enh = c(
            "WTonly_chr14:21460154-21460564",
            "WTonly_chr14:21297822-21298296"
          ),
          Relevance_int = c(21.32, 10.66),
          chr = c("chr14", "chr14"),
          start = c(21460154L, 21297822L),
          stop = c(21460564L, 21298296L),
          chr_trscpt = c("chr14", "chr14"),
          start_trscpt = c(21365694L,
                           21365694L),
          stop_trscpt = c(21366295L, 21366295L),
          Gene = c("Ppp3cb",
                   "Ppp3cb")
        ),
        .Names = c(
          "ID_trscpt",
          "ID_enh",
          "Relevance_int",
          "chr",
          "start",
          "stop",
          "chr_trscpt",
          "start_trscpt",
          "stop_trscpt",
          "Gene"
        ),
        row.names = c(7981L, 7984L),
        class = "data.frame"
      )


# construct a fake alignment file for sashimi plots. For each relevance point connection, add 10 fake aligned read pairs
read_weights <- floor(eppairs[,"Relevance_int"]*10)

sashimidata <- eppairs[,c("start","stop","start_trscpt","stop_trscpt")]
sashimidata <- apply(sashimidata,1,function(x){
  y <- sort(as.numeric(x))
  data.frame("start"=y[1],"stop"=y[4],"cigar"=paste0(y[2]-y[1],"M",y[3]-y[2],"N",y[4]-y[3],"M"))
  })

sashimidata <- data.frame("chr"=rep(eppairs[,"chr"],read_weights),do.call("rbind",sashimidata[rep(names(sashimidata),read_weights)]))


track_ep_pairs <- AlignmentsTrack(range=with(sashimidata,GRanges(chr,IRanges(start,stop),"*",cigar)),genome="mm9",isPaired=FALSE)

# works, I see gapped reads
plotTracks(list(track_ep_pairs),extend.left = 0.1, extend.right = 0.1)
# Error
plotTracks(list(track_ep_pairs),extend.left = 0.1, extend.right = 0.1,type="sashimi")

#Error in GAlignments(seqnames = seqnames(range), pos = start(range), cigar = range$cigar,  : 
#                        'cigar' must be a character vector with no NAs

# BUT:
#table(sashimidata$cigar)
#601M93859N410M 474M67398N601M 
#213            106 

#anyis.na(sashimidata$cigar))
#[1] FALSE

Answer 1

[Matthias Zepper]

Thanks to everybody, who read this post and tried to help. I could solve the problem now.

I was unfamiliar with the data format of Gviz tracks, but it turned out that the underlying data can be manipulated just as a regular GRanges object. After I finally found out that a simple

ranges(track_ep_pairs)

gives you access to the data used to generate a track, I could analyze it and track down issues. While mcols() won't work on tracks, ranges() does. It turns out that a simple factor to character conversion of the cigar column will do:

track_ep_pairs <- AlignmentsTrack(range=with(sashimidata,GRanges(chr,IRanges(start,stop),"*",cigar)),genome="mm9",isPaired=FALSE)
ranges(track_ep_pairs)$cigar <- as.character(ranges(track_ep_pairs)$cigar)

Now the sashimi plots render like a charm. Yeah!