Multiple Families for linkage map

[emma]

Hello,

This is a simple question but I couldn't find any information on it.
I am preparing to do a linkage map and I am going through the map
construction tutorial (which is great!). However, the tutorial has
only one family and I will have ~20 F2 families. I will also have
genotypes of the parents which I would like to put in there. How
should I code it in the csv file so that R/qtl reads it correctly?

-Emma

[Karl Broman]

If you have a number of intercrosses from inbred lines, you can tack them all together and load them as one cross, or you could load them separately and then combine them with the c() function. (Formally, it'd call the c.cross function.)

That parental genotypes are important suggests that these are crosses among outbreds. In that case they'd need to be treated as phase-known 4-way crosses. You need to figure out phase first (which is hard if you don't have a map.) And then you'd need to encode the genotypes in a special way (it's briefly explained in the help file for read.cross). Analysis of four-way crosses is not well documented and can be quite tricky. Unfortunately, R/qtl doesn't contain all of the necessary tools for constructing maps with outbred crosses.

karl

> --
> You received this message because you are subscribed to the Google Groups "R/qtl discussion" group.
> To post to this group, send email to rqtl...@googlegroups.com.
> To unsubscribe from this group, send email to rqtl-disc+...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/rqtl-disc?hl=en.
>

[emma]

I am pretty sure we don't have outbred crosses. We took individual
from two separate populations (geographically very separate) and used
them to make our mapping families as well as to do our RNA sequencing
for SNP discovery. So all SNPs we have are population specific.

I see your point about no bothering to genotype the parents since they
should be all heterozygotes anyways. Thanks for your help.
-Emma

On Feb 28, 1:54 pm, Karl Broman <kbro...@gmail.com> wrote:
> If you have a number of intercrosses from inbred lines, you can tack them all together and load them as one cross, or you could load them separately and then combine them with the c() function.  (Formally, it'd call the c.cross function.)
>
> That parental genotypes are important suggests that these are crosses among outbreds.  In that case they'd need to be treated as phase-known 4-way crosses.  You need to figure out phase first (which is hard if you don't have a map.)  And then you'd need to encode the genotypes in a special way (it's briefly explained in the help file for read.cross).  Analysis of four-way crosses is not well documented and can be quite tricky.  Unfortunately, R/qtl doesn't contain all of the necessary tools for constructing maps with outbred crosses.
>
> karl
>

[Karl Broman]

I wouldn't say "don't bother to genotype the parents". It's extremely important, for quality control, to have genotypes on the progenitors and the F1 parents. But in crosses among inbred lines, they won't be considered further and they don't get imported into R/qtl.

karl