I also had difficulty when first generating hexbin plots for high dimensional mass cytometry data a couple of years ago. Essentially, the approach of using hexbin() followed by plot() and then hexVP.abline() doesn't work. It's as if the viewport in which hexVP.abline() is attempting to place the line is not 'speaking' with the output from the plot() function (or something along these lines...). It's possible to place it after much hassle and time-wasting.

A better approach is to use a combination of hexbinplot() and a function from the lattice package. This will allow you to place the ablines with precision.

require(hexbin)
require(lattice)

require(RColorBrewer)
rf <- colorRampPalette(rev(brewer.pal(9,"BuPu")))

hexbinplot(CD28 ~ CD39, data=temp, aspect="1", xbins=500, cex.labels=1.0, cex.title=1.0, colramp=rf,
    panel=function(x, y, ...)
    {
        panel.hexbinplot(x, y, ...)
        panel.abline(v=c(0,4), h=c(0,4), col="black", lwd=4, lty=5)
    }
)

------------------------------------

Note that you can also set x- and y-axis limits, and fit a linear regression line:

rf <- colorRampPalette(rev(brewer.pal(9,"Spectral")))
hexbinplot(CD28 ~ CD39, data=temp, aspect="1", xbins=500, type="r", cex.labels=1.0, cex.title=1.0, colramp=rf, xlim=c(-3,8), ylim=c(-3,8))

-----------------------

Like the lattice package functions, which are great and flexible or plotting data in various ways, it's probably worth an entire day exploring hexbinplot() (if you can convince your supervisor...).

Kevin