Agriculture Reference
In-Depth Information
associations without normality. Once the detection of changes in shape
and shifts in the distribution are done, detecting signi
cant associations
becomes possible but it is necessary to examine the empirical distribu-
tion to con
rm the nature of the effect (Beló and Luck 2010).
E. Determining the Level and Influence of Population Structure
Population structure re
ects the history of a population. It could be due
to the selection of a particular haplotype or alleles during the history of
the species. The presence of PS can easily result in false positives during
AM. PS determination is done in terms of LD and allele distribution. PS
considers an unequal distribution of the alleles and some groups of
alleles tend to be associated due to PS but not physical linkage.
In structured populations, highly signi
cant but false-positive asso-
ciations between a marker and a phenotype may be observed, even
though that marker may not be physically linked to the locus responsible
for the phenotypic variation (Buckler and Thornsberry 2002). Determin-
ing the presence of PS will ensure that associations are real by correcting
for PS or determining where structure does not need to be considered. PS
should be determined with neutral alleles, because these markers are not
a focus of natural selection.
The effects of PS can also be corrected by using a large number of
independent genetic markers across the genome (Flint-Garcia et al.
2005). There are two major statistical methods to determine PS
—
genome
control (GC) and structure association.
1. Genome Control in Association Mapping.
To adjust for population
strati
control studies, Devlin and Roeder (1999) pro-
posed a statistic that accounts for the impact of substructure by using the
distribution of markers in the sampled genome (Devlin et al. 2001).
Hidden population structure will in
cation in case
-
ate the variance of the trend test.
The GC principle states that
“
if the variance of the candidate gene is
in
As a result, the
variance of the candidate gene can be corrected with the variance of the
null loci (Zheng et al. 2005).
The GC is correlated with
ated, the null loci variance will be in
ated too.
”
2
because it is in
, which
is proportional to the population structure present. The value of
ated by a factor of
χ
λ
is
estimated by examining random loci across the genome and then
this should be incorporated into the test by rescaling the chi-square
statistic (Devlin and Roeder 1999). Alternatively, the use of the Cochran
λ
-
