Does anybody know of any work regarding the potential bias of an
estimated linkage map, when members of the segregating cross are
selectively genotyped with respect to a quantitative trait. It
seems to us that in regions without QTLs, there will be absolutely
no bias, while in regions containing QTLs, any bias would be
proportional to the magnitude of QTL effects. If selective
genotyping was directional, then an excess of markers of the
favored parent would be found.
For example, suppose large and small flowered taxa were crossed,
and their F1 backcrossed to large flowered plants. When large
flowered plants are selectively genotyped in this BC generation,
then the 1:1 ratio would be distorted in the direction of the
allele possesed by the large flowered taxa. But even if the ratio
was distorted, how much bias of linkage distances and order would
occur? Would we expect much in the case of QTLs of mild effect?
In our case, we selectively genotyped equally in both directions.
(e.g. the upper 25% and lower 25% of the flower size distribution
was genotyped, then the map constructed). We think that even in
regions with QTLs of major effect, the bias would be small or
non-existant, as opposing the directions of selection cancel (one
would now get a 1:1 ratio despite the selection).
We am posing this question due to a reviewers comment on our RAPD
map in Mimulus (monkeyflowers). Some work has been done on the
effect of directional natural selection on mapping QTLs (Mackinnon
and Georges, Genetics 132: 1177-1195, 1992), but we cannot find
references the effects of directional or disruptive selection on
construction of the genetic map itself.
Kermit Ritland and Jingzong Lin
ritland at botany.utoronto.ca, lin at botany.utoronto.ca
Department of Botany
University of Toronto
