Software

Some Notes On The Marker Regression For Qtl Mapping Matlab Code:

The entry point is 'script01.m' which walks through a typical analysis.
everything is based on marker regression for which 10cM marker spacing
is ideal.  Less is typical however and we go ahead with it.  there is an
imputation step that fills in missing marker genotypes.  It is valid for
genotype data that have a few holes.  If you've only typed extremes or
have a lot of missing data that will require more care.

The first step of the script is to run 'data_in.m' This will have to be
modified slighly to your needs, but it bascially expects to find the
following files

geno.dat  a file of genotypes coded as 0-1 or 0-1-2
        with 9 for missing data.  mice are rows.
        markers are columns.  this reversed from the
        mapmaker convention.  markers should be in
        order by chromosome number and along chromosomes.

mnames.txt  a list of marker names, in order, one name per row.
        there is a bit of kludge in the data_in script to
        avoid getting a blank name [] at the end of the list.

chrid .dat a column of chromosome numbers corresponding to
        mnames.txt

pheno.dat  a file of phenotypes.  mice are rows.  distinct
        phenotypes are columns.  modify if only one phenotype.
        missing data code can be whatever...i used -999 here.
        missing data complicate the analysis script somewhat.

After data_in gets run, some checking of dimensions is in order.  use
'size()' in matlab to see that the data objects (geno, genoi, chrid,
mnames and pheno) all have conformable dimensions.  To be off-by-one is
not uncommon and needs to be fixed.

The rest of the script runs through a genome scan for main effects and a
genome scan for joint effects of marker pairs (including a test for
interaction).  The permutation test for the latter is time consuming and
may be commented out.  We find that F > 5 is about right as genomewide
cutoff for pairwise analysis of F2 populations.  Try the permutation
test with 10, then crank it up to 100 or 1000 and go home for the night.

Source code can (and should) be modified as needed for specific
applications.  We can answer questions. The software is intended for
sophistacted users.  There are no immediate plans (funds) for user
friendly software development.