A simple regression strategy for mapping multiple linked
quantitative trait loci (QTLs) in inbred
populations is proposed and applied to data from a non-obese diabetic (NOD)
mouse backcross.
The method involves adding and deleting markers from a linear model in
a stepwise manner,
allowing the association with a particular marker to be examined once
associations with other (in
particular neighbouring) markers have been taken into account. This approach
has the advantage
of using programs available in standard statistical packages while still
allowing adequate
separation of possible multiple linked effects. For the mouse backcross,
using these methods, at
least two and possibly three diabetogenic loci are detected on each of
chromosomes 1 and 3. Some
evidence for epistasis is seen between the loci on chromosome 1, with a
possible additional
epistatic interaction between the loci on chromosome 3. Congenic strain
analysis of the
chromosome regions in NOD diabetes suggests that although the true type
I error rate may be
larger than that suggested by the nominal P values, our
results nevertheless correspond well with
those disease loci and interactions detected using a congenic approach,
indicating that the
regression method may be a powerful strategy for the detection and characterization
of QTLs in
inbred populations.