Agriculture Reference
In-Depth Information
As you can see, if there is high positive correlation
between two variates it will be possible to select individ-
ual genotypes which have high expression in both traits.
Conversely, if there is high negative correlation between
variates it may be very difficult to select genotypes with
high expression in both characters. The magnitude of
the correlation value, in absolute terms, can be asso-
ciated with underlying physiological processes or even
pleiotropy (i.e. the same genes directly control expres-
sion in the two characters) or is a reflection of genetic
linkage.
It would seem obvious that there must be some rela-
tionship between
r
, the correlation coefficient, and
h
2
,
the heritability. If therefore characters are recorded on
the same set of genotypes grown in two different envi-
ronments (say locations or years), then the magnitude
of the correlation coefficient indicates the relationship
between performances in the different environments.
Squaring the correlation coefficient (
r
2
) provides an
estimate of the proportion of the total variation between
the environments that is explained by the correlation.
The total variation between sites can be considered
the total phenotypic variation, and as the proportion
accounted for by regression must have a genetic base,
then a simple relationship exists, whereby
r
2
However, it is very important that the limitations
of heritability estimates are realized. These limitations
occur on at least three levels:
•
There are many technical assumptions inherent in
the theory as presented (e.g. that genes assort inde-
pendently, that alleles segregate independently, that
there is no epistasis). It is possible to allow for many
of these complications, but only at the expense of
making the theory more complicated
•
An estimate of narrow-sense heritability strictly
applies to a
particular
character, in a
particular
pop-
ulation, at a
particular
moment and in a
particular
environment. Thus, even for a single character, her-
itability is not constant. It is obvious that
h
2
is
particularly vulnerable to changes in environmental
variance, changes that can occur at the same place
at different times, different places at the same time,
or both. But widely different estimates of
h
n
for the
same character can be found in different populations
investigated at the same place and time. Also, it has
been seen that additive genetic variance, and hence
narrow-sense heritability, generally declines over gen-
erations of selection, even in a constant environment.
Caution must therefore be exercised in interpreting
estimates of
h
n
if it is not known that every precau-
tion has been taken to expose different populations
and/or characters to the same range of environments
and one is interested only in the response in the same,
or very similar, environments
is a direct
estimate of
h
2
.
Heritability and its limitations
In this short but important section, a critical look is
taken at the concept of heritability, its uses and misuses.
Four distinct methods of estimating narrow-sense
heritabilities have been outlined:
•
While means are what are called first degree statis-
tics, variances, etc. are second degree. Second degree
statistics are usually 'less precise' than first degree.
h
n
and
h
b
(with the exception of
h
n
from mid-parent
onto offspring regression) being based on the ratios
of variances, share all the weaknesses of second degree
statistics.
•
Partitioning of phenotypic variances
Offspring-parent regression
•
Response to selection
•
Correlation
•
Methods of selection
How response to selection can be predicted from a
given selection differential when the narrow-sense her-
itability is already known from other experiments, has
also been covered above. The concept of heritability, and
estimates of it, have been of great value to plant breeders
and to population geneticists interested in continuously
varying characters in natural populations.
When a plant breeder is selecting a particular population
for only a single trait the operation is usually relatively
simple. The population is evaluated for the character in
question and those phenotypes with desirable (whether
this is high, low or intermediate) expression are
selected
while the phenotypes with less desirable expression are
