S7 compartmap objects for analysis

[jamespeapen]

S7 objects for analysis

Creates a CompartmentCall S7 object with a vector of inferred singular values from compartmap or PC1 from Hi-C, the genomic resolution and a GRanges of the bins:

<compartmap::CompartmentCall> class
@ parent     : <S7_object>
@ constructor: function(pc, res, gr, name, unitarized) {...}
@ validator  : <NULL>
@ properties :
 $ name      : <character>
 $ gr        : S4<GRanges>
 $ dt        : S3<data.table>
 $ res       : <integer> or <double>
 $ unitarized: <logical>

• @name: an identifier for the object
• @res: the resolution at which compartments were called
• @gr: a GRanges object with the bins
• @dt: a data.table with index n, singuar values pc, and the name.
  This is used to store the values and to plot the object
• @unitarized: Whether the values have been unitarized

CompartmentCall implements

• [ return a new CompartmapCall object with subset data and granges
• plot: line plot of the singular values
• flip: return a new CompartmentCall with sign-flipped singular values
• unitarize: return a new CompartmentCall with unitarized singular values
• DF: get the @dt slot
• granges: get the @gr slot
• resolution: get the resolution
• is_unitarized: get unitarized state
• get_name: get object name
• print: print the object and property metadata

CompartmapCall inherits from CompartmentCall but takes the GRanges output of scCompartments() to construct itself:

<compartmap::CompartmapCall> class
@ parent     : <compartmap::CompartmentCall>
@ constructor: function(gr, res, name, unitarized) {...}
@ validator  : <NULL>
@ properties :
 $ name      : <character>
 $ gr        : S4<GRanges>
 $ dt        : S3<data.table>
 $ res       : <integer> or <double>
 $ unitarized: <logical>

MultiCompartmentCall also inherits from CompartmentCall but takes a list of CompartmentCall and/or CompartmapCall objects. It checks that all input GRanges, and resolutions are the same and that they are all either unitarized or non-unitarized, unless users want to unitarize inputs with unitarize=TRUE.

<compartmap::MultiCompartmentCall> class
@ parent     : <compartmap::CompartmentCall>
@ constructor: function(ccalls, name, unitarized, unitarize) {...}
@ validator  : function(self) {...}
@ properties :
 $ name      : <character>
 $ gr        : S4<GRanges>
 $ dt        : S3<data.table>
 $ res       : <integer> or <double>
 $ unitarized: <logical>
 $ colnames  : <character>
 $ mat       : S3<matrix/array>

MultiCompartmentCall implements

• [: returns a new MultiCompartmentCall object with both row and column subset
• print: calls CompartmentCall's print method and adds @mat dimensions
• corr: gets correlations of every pair of input singular value vectors
• agr: gets agreement of every pair of input singular value vectors
• plot: plots singular values by color based on the inputs CompartmentCall
name property

To compare the example k562 calls at 100kb with the HiC PC1:

example_100kb <- CompartmapCall(
  example_100kb.gr, # the GRanges from scCompartments
  res = 1e5,
  name = "example_100kb_binary"
) |> unitarize()

hic_100kb <- CompartmentCall(
  pc = na.omit(scan(hic_reference$kb100)), # text file containing PC1
  res = 1e5,
  gr = granges(example_100kb.gr), # uses the GRanges from the scCompartments
  name = "hic_100kb",
  unitarized = TRUE
)

example_hic_100kb <- MultiCompartmentCall(
  c(example_100kb, hic_100kb),
  name = "example_hc_100kb"
)

corr(example_hic_100kb)

#                      example_100kb_binary hic_100kb
# example_100kb_binary            1.0000000 0.6347439
# hic_100kb                       0.6347439 1.0000000

agr(example_hic_100kb)
#                      example_100kb_binary hic_100kb
# example_100kb_binary            1.0000000 0.7485844
# hic_100kb                       0.7485844 1.0000000

plot(example_hic_100kb)

I chose S7 because its simpler than the S4 system, but it does mean that the subsetting and print methods need to be written from scratch.