manipulate GenomicInteractions object

[mjko1210]

Hi,

I'm wondering if I can modify GenomicInteractions object.

For example,
data <- makeGenomicInteractionsFromFile(f, type="bedpe")

GenomicInteractions object with 65 interactions and 1 metadata column:
seqnames1 ranges1 seqnames2 ranges2 | counts
|
[1] chr13 63039538-63039537 --- chr17 22147340-22147339 | 10
[2] chr6 76245070-76245069 --- chr13 65568744-65568743 | 10
[3] chr13 65720196-65720195 --- chr21 9851052-9851051 | 10
[4] chr13 60977697-60977696 --- chr22 46771901-46771900 | 10
[5] chr6 76245101-76245100 --- chr13 65568774-65568773 | 10
... ... ... ... ... ... . ...
[61] chr13 49147000-49146999 --- chr13 66195900-66195899 | 10
[62] chr13 59255135-59255134 --- chr13 59476455-59476454 | 10
[63] chr13 59263605-59263604 --- chr13 63699832-63699831 | 10
[64] chr13 59475210-59475209 --- chr13 61920391-61920390 | 10
[65] chr13 63043465-63043464 --- chr17 22143270-22143269 | 10

I was able to access counts via data$counts, however not for any other columns (such as seqnames and ranges).

I've tried converting object to data frame using as.data.frame(data). However, I'm not sure how to convert it back to GenomicInteractions object after the modification.

Your advice will be appreciated!

Thanks!

[liz-is]

Hi,

The GenomicInteractions object stores region information as a GRanges object, along with information about which regions are interacting with each other, and counts and other metadata stored in metadata columns. You can read more about how the object is structured in the vignette, or the InteractionSet vignette (as we inherit from InteractionSet objects), and that might help understand how to manipulate it.

We don't recommend modifying a GenomicInteractions object in place, but if you want to create a new GenomicInteractions object I would probably suggest using this approach:

new_gi <- GenomicInteractions(anchor_one_granges, anchor_two_granges, counts, metadata)

You can easily create GRanges from a data frame using the makeGRangesfromDataFrame function, changing the start.field and end.field arguments as appropriate if you're storing information for more than one genome region in the same data frame.

[mjko1210]

Thank you for your comments.
I've tried making GenomicInteractions object following your advice.
It works fine with anchor_one_granges, anchor_two_granges, counts but not with metadata.

anchor_one = data[,c(1:3,9)]
anchor_two = data[,c(4:6,10)]
colnames(anchor_one) = c('chr','start','end','strand')
colnames(anchor_two) = c('chr','start','end','strand')

anchor_one_granges = makeGRangesFromDataFrame(anchor_one)
anchor_two_granges = makeGRangesFromDataFrame(anchor_two)
counts = data[,8]
metadata = data[,11] #contains types
new_gi <- GenomicInteractions(anchor_one_granges, anchor_two_granges, counts, metadata)

Error in (function (classes, fdef, mtable) :
unable to find an inherited method for function 'GInteractions' for signature '"GRanges", "GRanges", "character"

I've also tried by making both counts and types as metadata

metadata = DataFrame(counts = data[,8], types = data[,11])

However, I'm still getting error:

new_gi <- GenomicInteractions(anchor_one_granges, anchor_two_granges, metadata)
Error in (function (classes, fdef, mtable) :
unable to find an inherited method for function 'GInteractions' for signature '"GRanges", "GRanges", "DFrame"'

I'm not sure how other column can be added as metadata except counts here.

Thanks!

[liz-is]

Error in (function (classes, fdef, mtable) :
unable to find an inherited method for function 'GInteractions' for signature '"GRanges", "GRanges", "character"

Is your "counts" column encoded as character by any chance? try as.numeric(counts) first. You can see all the possible ways of constructing a GenomicInteractions object by looking at ?GenomicInteractions.

If that doesn't solve it, please provide a reproducible example with an example BEDPE file and/or an example of what your data object looks like, e.g. by using dput().