ENU: combining backcross and F2

[Joanne]

Hello,

Is it possible to combine backcross data with F2 data - assuming the
parental strains are the same, and all mice are genotyped at the same
markers? Is this likely to affect qtl mapping?

The reason I'm asking, is that I am currently analyzing an ENU
pedigree for resistance to infection (background strains are
susceptible). We originally set it up as a backcross so that that the
G2 daughters (hopefully carrying a useful mutation - heterozygote) are
crossed back to their G1 father (heterozygote) to produce the N3
generation wherein approximately 25% of offspring will carry a
homozygous mutation. N3s are phenotyped and genotyped.

We used this approach to identify an interesting pedigree, but we were
unable to genotype enough N3 mice to map the locus. Rather than lose
this mutation, we intercrossed within the G2 males and females (ie.
brother-sister rather than father-daughter), hoping to match a
heterozygous mutant male with a heterozygous mutant female. Ten
random couples were set up and we used progeny testing on these F2s to
identify which couples were carrying the mutation.

Now I want to genotype this F2 generation and combine it with the
previous N3 generation to map the qtl.

Can I combine these data sets just by adding more mice to the list, or
do I need do some more complex adjustments

thanks!
Joanne

I hope that made sense - I would be happy to email a drawing of the
pedigree set-up to clarify things, but I couldn't find a way to attach
it in this post.

[Karl Broman]

Dear Joanne,

It is possible to analyze combined backcross/intercross data with R/
qtl, though it can be a bit tricky, I've not tested this thoroughly,
and it may require a bit of programming and manipulation of the data
objects.

R/qtl doesn't handle N3 data exactly right, but if you have reasonably
dense and complete marker data, analysis as if it were a regular
backcross should be okay.

You want to load the data, do genotype probability calculations (with
calc.genoprob), and then combine. That's so that the genotype
probabilities are calculated appropriate for the individuals; if you
combined first, the backcross individuals would be treated the same as
the intercross individuals, which isn't quite right.

There's some code below that illustrates some of this with simulated
data.

karl

-----------

# simulate data with a single QTL; just the autosomes
data(map10)
bc <- sim.cross(map10[1:19], model=c(3, 15, 0.5), type="bc", n.ind=200)
f2 <- sim.cross(map10[1:19], model=c(3, 15, 0.5, 0), type="f2",
n.ind=200)

# calculate genotype probabilities
bc <- calc.genoprob(bc, step=1)
f2 <- calc.genoprob(f2, step=1)

# separate analyses
out.bc <- scanone(bc)
out.f2 <- scanone(f2)

# combined data
comb <- c(bc, f2)
thecross <- pull.pheno(comb, "cross")
out.comb <- scanone(comb, addcovar=thecross)

# plot the results, just for chr 3
plot(out.comb, out.bc, out.f2, chr=3)
legend("topright", c("Combined", "Backcross", "Intercross"),
col=c("black","blue","red"), lwd=2)