Agriculture Reference
In-Depth Information
stituting one allele at a locus for another and can
be estimated through measures of the level of
inbreeding and the genetic relationship among
sibs via the covariance among relatives (Falconer
and Mackay 1996). Where heritability is low,
alternate populations containing greater additive
genetic variance (and high mean) should be con-
sidered, or sampling should be modifi ed to reduce
nongenetic variation.
Alternatively, surrogate traits may occur for
characteristics that are costly or diffi cult to
measure such as aerial biomass or grain yield. If
the additive genetic correlation (
r
A
) between two
traits (X and Y), and their narrow-sense herita-
bilities
h
2
X
and
h
2
Y
are known, the correlated
response of trait Y to selection on trait X (D
GY.X
)
can be predicted by:
potentially useful for implementation in a breeding
program: (i) increased genetic variability; (ii) high
heritability; (iii) opportunity for out-of-season
selection; (iv) simplicity and lower cost relative to
yield selection; (v) applicability to marker-assisted
selection; (vi) an allowing for targeted selection of
traits and therefore genes from unadapted donors;
and (vii) the capacity to be assessed singly or with
other traits in simulation modeling to determine
their value to breeding (after Richards 2006). Yet
despite this potential there are few examples
where physiological traits have been successfully
implemented in breeding programs (Jackson
et al., 1996; Richards 2006). Lack of adoption may
refl ect the following: (i) trait type (survival and
not productivity based); (ii) low heritability and/
or low genetic correlation with yield; (iii)
environment-specifi c expression; (iv) repulsion-
phase linkages for the target gene and other
important genes (especially in wild donors);
and (v) high cost of phenotyping. Finally,
implementation requires good communication for
integration from the physiologist and molecular
biologist to the breeder.
Δ
GY.X
=
k
s
pY
h
Y
h
X
r
A
,
(11.1)
in which
k
is the standardized selection differen-
tial, and s
pY
is the phenotypic standard deviation
for trait Y (Falconer and Mackay 1996). If
h
Y
<
h
X
r
A
, then selection for trait X will result in greater
change in trait Y than direct selection for Y. As a
proxy for
r
A
,
the genetic correlation for two vari-
ables can be readily estimated from analysis of
covariance. Linkage disequilibrium (population
type), chromosomal linkage, and pleiotropy can
all contribute toward two traits being genetically
correlated. However, owing to recombination,
only pleiotropic effects are likely to maintain a
genetic association over cycles of crossing and
selection in a breeding program. Most studies
report only phenotypic correlations, which differ
from genetic correlations as phenotypic correla-
tions also contain environmental and sampling
covar iance components (Searle 1961). Importantly,
only the genetic covariance component of this
correlation (more specifi cally, the additive
correlation) is responsive to selection.
Several traits have been reported as under
pleiotropic control with other traits at one or
more loci. For example, carbon isotope discrimi-
nation (CID) measured prior to anthesis has
shown a strong additive genetic correlation with
grain yield and biomass (Rebetzke et al., 2002).
Indirect selection for yield and biomass is more
BREEDING TOOLS
Indirect selection via correlated traits
Breeding traditionally employs direct selection for
genetic gain of target traits. However, the success
of direct selection is contingent on populations
containing adequate additive genetic variance and
high narrow-sense heritability. The trait should
also be simple and inexpensive to measure, par-
ticularly in mass selection where many families
may be evaluated. Narrow-sense heritability (
h
2
)
can be calculated on a line-mean basis as follows:
h
line-mean
= s
A
/ (s
A
+ s
AE
/
n
e
+ s
residual
/
n
r
n
e
),
in which s
2
A
, s
AE
, and s
2
residual
are estimates of the
additive, additive × environment, and residual
variances, respectively, and
n
e
and
n
r
are the
number of environments and replications per
environment, respectively. The additive genetic
variance provides a measure of the effect of sub-
