Visualizing CNV data based from bed or bedgraph format

Question

[sm.hashemin]

Hi,

I was wondering if there is a package or script which may visualize multi-panel CNV just IGV does but with more flexibility from a cohort of bedgraph files? preferably in R? Sushi R package does not have a point based visualization otherwise might have been perfect.

Best

[Arup Ghosh]

Have you checked aCNViewer tool?

Answer 1

[bernatgel]

Hi sm.hasehmin,

You can use the Bioconductor package karyoploteR for this. You can plot the data as points using kpPoints as in the example below or you could change that to kpSegments to show the width of your data elements, for example. There's more information on the available functions and customization options at the karyoploteR tutorial.

To load the bedgraphs into R you can use toGRanges from regioneR. In the example I'm using bedgraph example data from Sushi which is not CNV data, but with only minor changes (setting ymin and ymax) you should be able to adapt the code to CNV values. To load your data, you can give toGRanges the file name of your bedgraphs and it should load them.

For the vertical positioning (r0 and r1) I'm using autotrack. This function is not yet in the tutorial but the inline documentation should suffice.

Note: since the example datasets have only data for a small region of chromosome 11 this is what we are plotting. Removing the zoom argument from plotKaryotype will plot the whole genome.

#Load 3 example bedgraphs from the datasets in the Sushi package
library(Sushi)
data(list=c("Sushi_DNaseI.bedgraph","Sushi_ChIPSeq_CTCF.bedgraph",
        "Sushi_ChIPExo_CTCF.bedgraph"), package="Sushi")

dd <- list(CTCF_Exo=Sushi_ChIPExo_CTCF.bedgraph,
           CTCF_Chip=Sushi_ChIPSeq_CTCF.bedgraph,
           DNase=Sushi_DNaseI.bedgraph)

png("bedgraphPlot.png", width = 1500, height = 1000)
  kp <- plotKaryotype(zoom="chr11:1643216-2359707", main = "BedGraphs", cex=3)
  kpAddBaseNumbers(kp, tick.dist = 10e4, add.units = TRUE, cex=1.8)
  for(i in seq_len(length(dd))) {
    names(dd[[i]]) <- c("chr", "start", "end", "value") 
    gr <- toGRanges(dd[[i]])
    at <- autotrack(i, length(dd), margin = 0.1)
    kpPoints(kp, data=gr, ymax=max(gr$value), r0=at$r0, r1=at$r1, col=rainbow(10)[i])
    kpAddLabels(kp, labels = names(dd)[i], r0=at$r0, r1=at$r1, cex=2)
  }
dev.off()

And you would get something like this

[sm.hashemin]

Thanks alot. I removed the zoom argument and the representation was a bit two detailed for an overview of genomic changes. Is the reperesentaion like this possible ? https://us.v-cdn.net/5019796/uploads/FileUpload/58/15ca12b0618250b498391fe5562fb9.png

[bernatgel]

Sure, to have all chromosomes in one line use

kp <- plotKaryotype(plot.type=4)

and keep everything else the same. There are various plot types available to choose from. You can find them in the tutorial page.

The SNP-array data example might be useful to you. At the end your data is very similar to the LRR in SNP arrays.

Answer 2

[Pierre Lindenbaum]

some tools I wrote:

http://lindenb.github.io/jvarkit/WesCnvSvg.html

It's CNV friday. (found with @brent_p 's indexcov) pic.twitter.com/40maSGmCzg

— Pierre Lindenbaum (@yokofakun) September 14, 2018

http://lindenb.github.io/jvarkit/WesCnvTView.html

a first screenshot: pic.twitter.com/1qUkcZ80Df

— Pierre Lindenbaum (@yokofakun) October 19, 2018

Answer 3

[sm.hashemin]

Thanks a lot for the answers I am more searching for a representation like the following. As I have no normal samples I had to use ControlFREEC and reanalysis with another software is not an option.