Height values for each point in a plot

问题

I have a data of protein-protein interactions in a data frame entitled: s1m. Each DB and AD pair make an interaction and I can plot it as well:

> head(s1m)
     DB_num AD_num
[1,]      2   8153
[2,]      7   3553
[3,]      8   4812
[4,]     13   7838
[5,]     24   3315
[6,]     24   6012

Plot of the data looks like:

I then used code I found on this site to plot filled contour lines:

## compute 2D kernel density, see MASS book, pp. 130-131
require(MASS)
z <- kde2d(s1m[,1], s1m[,2], n=50)
plot(s1m, xlab="X label", ylab="Y label", pch=19, cex=.4)
filled.contour(z, drawlabels=FALSE, add=TRUE)

It gave me the resulting image(minus the scribbles): result

MY QUESTION: I need to annotate each line of data in the original s1m data frame with a number corresponding to its height on the contour map (hence my scribbles on the image above). I think the list z has the values I am looking for, but I am not sure.

In the end I would want my data to hopefully look something like this so I could study the protein interactions in groups:

         DB_num AD_num   height
    [1,]      2   8153        1
    [2,]      7   3553        1
    [3,]      8   4812        3
    [4,]     13   7838        6
    [5,]     24   3315        2
    [6,]     24   6012        etc.

回答1:

This is one option if you want the actual height not the bin each is assigned to

## dummy data
DF <- data.frame(DB_num = rnorm(10000), AD_num = rnorm(10000))

require("MASS")

kde <- kde2d(DF[,1], DF[,2], n = 50)

Note the kde2d returns as component z which is a matrix with (in this case) 50 rows and columns where rows correspond to the x data and columns to the y data. As a matrix is just a vector, and the data are filled by columns, we can exploit this and stack the x and y values n times each (n = 50 here), then unwind kde$z

dd <- dim(kde$z)
res <- data.frame(DB_num = rep(kde$x, times = dd[1]),
                  AD_num = rep(kde$y, times = dd[2]),
                  height = as.numeric(kde$z))

This produces

> head(res)
        DB_num      AD_num                                  height
1 -3.582508378 -3.79074271 0.0000000000000000000000000006907447484
2 -3.429230262 -3.63682706 0.0000000000000000000000002951259863229
3 -3.275952146 -3.48291141 0.0000000000000000000000558203373144190
4 -3.122674029 -3.32899576 0.0000000000000000000055565720524140235
5 -2.969395913 -3.17508011 0.0000000000000000014967010810961022503
6 -2.816117797 -3.02116446 0.0000000000000008159370528768207499471

To get the bins, you'd need to follow what filled.contour did, which is to form breaks via

nlevels <- 20 ## default
brks <- pretty(range(res$height), nlevels)

> brks
 [1] 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.10 0.11 0.12 0.13 0.14
[16] 0.15 0.16

Then use cut to assign each height to a bin on basis of brks, something like

res <- transform(res, bin = as.numeric(cut(height, brks)))

Which gives

> head(res)
        DB_num      AD_num                                  height bin
1 -3.582508378 -3.79074271 0.0000000000000000000000000006907447484   1
2 -3.429230262 -3.63682706 0.0000000000000000000000002951259863229   1
3 -3.275952146 -3.48291141 0.0000000000000000000000558203373144190   1
4 -3.122674029 -3.32899576 0.0000000000000000000055565720524140235   1
5 -2.969395913 -3.17508011 0.0000000000000000014967010810961022503   1
6 -2.816117797 -3.02116446 0.0000000000000008159370528768207499471   1

You'll probably want to check the details of ?cut to determine behaviour on the boundary of a bin, but that should get you close enough.

来源：https://stackoverflow.com/questions/17795204/height-values-for-each-point-in-a-plot

标签

data-visualization

scatter-plot

kernel-density