Heritability

• Quantitative traits are controlled by a large number of

genes with additive effects.
• The phenotypic value is used as an indirect measure of the
effects of the genes, with consideration of both genotypic
and environmental influences.
• The total genetic variance VG = VA + VD + VI
• VA is the additive genetic variance, VD is the variance due
to dominance effects, and VI is the variance due to
epistatic effects.
• Fish that are superior because of VA are able to transmit
their additive effects and contribute to progress in
selection programme.
• If you know the VA for a quantitative phenotype, you
can predict the phenotypic mean in the next
generation, based on the mean of the selected
broodstock, thus VA is also called the variance of
breeding value.
• Breeding values (additive genetic merit of a fish) are
of prime importance in selection programmes,
because the main aim when deciding which fish to
use as breeders is to choose those which are going to
have the best offspring.
• The breeding value is defined as the effect of an
individual’s genes on the value of the trait in its
offspring; this effect is caused by the additive effect
of genes.
• In breeding programme it is important to know the
proportionate amount of VA so that you can predict
whether selection will be effective.
• The proportionate amount of phenotypic variation
(VP) that is attributable to additive genetic variation
(VA) is called heritability (h2).
• h2 = VA/VP
• Heritability describes the percentage of VP that is
inherited in a predictable and reliable manner.
• VA/VP is called heritability in the narrow sense, in
contrast to the ratio VG/VP, which is heritability in
the broad sense, and which describes the relative
contribution of genotypic variance to VP
• Since VA is never greater than VG, then narrow-
sense heritability is never greater than broad-sense
• Because the concept of heritability is most
commonly used in relation to selection
programmes, and because it is breeding values that
are of prime importance in selection programmes,
the term heritability is normally used in the
narrow sense.
• Once you know the heritability for a quantitative
phenotype, you can predict the response to
selection by using the following formula: R = h2S,
where R is response (gain or loss) per generation,
S is the selection differential.
• Response to selection (R) equals the mean
phenotypic value of the offspring population
minus the mean phenotypic value of the whole
population from which the parents were selected.
• Selection differential refers to the superiority or
inferiority of the selected broodstock over the
population average. It is equal to the mean
phenotypic value of the candidates selected as
parents minus the mean phenotypic value of the
whole population from which they were selected.
• From the equation R = h2S, it follows that, if h2 =
1, R = S; if h2 = 0, R = 0
• A channel catfish farmer decides to initiate a
selection programme for increased growth in his
farm, in which the catfish currently averages 454 g
at 18 months of age. To implement his
programme, the farmer selects 50 females that
average 604 g and 40 males that average 692 g. If
h2 for growth is 0.5, calculate the predicted
response to selection. What is the predicted
average weight in the next generation?
• First calculate the selection differential as follows:
• S = (mean wt of selected female + mean wt of selected male)/2 –
population mean
• S = (604 g + 692 g)/2 – 454 g = 194 g
• Then calculate the predicted response to selection
• R = h2S = 0.5*194 g = 97 g.
• The next generation of channel catfish should average 97 g
more than the parental generation, thus F 1 = parental
population average + response to selection
• F1 = 454 g + 97 g = 551 g.
• The actual gain can be greater or less than 97 g. Deviation
from the predicted response are due to VD and VI effects,
changes in the environment (VE or VG*E), and can also be
due to sampling error as a result of working with a small
population.
• Phenotypes with h2  0.25 can be changed
effectively by selection, Phenotypes with h 2 
0.15 are not easy to change by selection.
• The larger the h2, the easier it will be to change a
population mean by selection.
• Heritability can change because of changes in the
environment. Heritability is the proportionate
amount of VA, so any factor that changes the
denominator will change.
• h2 = VA/VA + VD + VI + VE + VG*E
• Heritability is not fixed for a trait, hence, there is
no universal heritability for a trait. The heritability
value for a trait depends upon the genetic makeup
and the specific environment of the population.
• Heritability value can be applied only to a
specific group of individuals in a specific
environment.
• Changing groups or environments does not
alter the way in which genes affect the trait,
but it may change the amount of genetic
and environmental variance for the trait,
which could then alter the heritability.
• On a practical level , this means that h 2
calculated in population other than the one
at your station should be used only as
guidelines.
 Estimation of heritability
• It is not mandatory to estimate h2 prior to
selection, but it is often advisable because it will
enable you to predict whether a breeding
programme is likely to be worthwhile.
• The basis of methods for estimating heritability is
resemblance between relatives. The most common
classes of relatives are full-sibs, half-sibs or
parents and their offspring.
• The main methods that are commonly used to
calculate h2 are: the regression offspring on
parent, the correlation between paternal half-sibs,
sib analysis and realized h2.
 The regression of offspring on parent
• In this method the phenotypes of parents and
offspring are measured in a series of families, and
then the relationship between their phenotypes is
analyzed.
• One method is mid-parent-offspring regression. In
this method the average value of offspring from each
family is paired with the average of the two parents.
• When the mean phenotype of the offspring is
regressed against the mean phenotype of the parents,
the narrow-sense h2 equals the slope of the regression
line (b)
• h2 = bӮ/X, where x̅ is the mean phenotype of the
parents and Ӯ is the mean phenotype of the offspring.
• Another method is parent-offspring regression.
• In this method the average value of the progeny
from each of the series of families is paired with
either the father’s or the mother’s values and a
regression of the paired data is used to determine
h2 .

• When the mean phenotype of the offspring is

regressed against the phenotype of only one
parent, the narrow-sense h2 is twice the slope of
the regression line (b)
• h2 = 2bӮ/ X , where x is the phenotype of one
parent and Ӯ is the mean phenotype of the
offspring.
• This is because an offspring shares only half of its
genes with one of the parent.
The correlation between paternal half-sibs
• In half-sib analysis, each male is mated to two or
more females to produce paternal half-sib
families.
• The correlation between the paternal half-sibs is
estimated as intra-class correlation (t), which
measures the similarity among observations
belonging to the same class.
• t = s2b/(s2b + s2w), where s2b is the variance between
progeny groups sired by different sires; and s 2w is
the variance between progenies of the same sire.
• t = ¼VA/VP , then h2 = 4t
• To compute s2b and s2w an analysis of variance is
carried out as follows:-
1. Compute total sum of squares and sum of
squares between sires.
2. The difference between total sum of squares and
sum of squares between sires is the sum of
squares within sires.
3. Compute total degrees of freedom (number of
observation less 1) and degrees of freedom
between sires (number of sires less 1). The
difference is the degrees of freedom within sires.
4. Compute mean squares between sires and within
sires by dividing respectively, the sum of
squares between and within sires by the
corresponding number of degrees of freedom.
5. The component of variance between sires is
estimated as follows:
s2b = MSQb – MSQw/n, where MSQb and
MSQw are mean squares between and
within sires, respectively, and n is the
number of observations per sire.
6. The component of variance within sires is
estimated as the mean square within sire;
s2w = MSQw
7. The intra-class correlation , which
measures the similarity among observations
belonging to the same class is calculated as
follows: t = s2b/s2b + s2w
 Example
In a breeding programme 6 sires were selected and
mated to random dams of Nile Tilapia to produce
offspring of the next generation. From the offspring
of each sire a random sample of 5 offspring were
measured for body weight at six months of age and
weight gain calculate. The weight gain of each
offspring are shown in the Table below. Using the
data indicated in the Table estimate the within-
progeny variance component (s2W), the between-sire
component (s2b), the intra-class correlation (t) and
heritability.
 Weight gain (g) of offspring of different sires

Sire
Offspring A B C D E F
1 46 70 52 47 42 35
2 31 59 44 39 64 68
3 37 64 57 70 50 59
4 62 70 40 46 69 38
5 30 57 67 60 77 57
CF = G2/N = 16072/30 = 86081.63

Total sum of squares = 462 +312 + 372 + …+ 572 – CF

= 91213.0 – 86081.63 = 5131.37

Between sire sum of squares = (2062 + 3202 + 2602 +

2622 + 3022 + 2572)/5 – CF
= 438333/5 – 86081.63 = 1584.97

Within sire sum of squares = 5131.37 – 1584.97 =

3546.40
Source of df SS MS Variance components
variation
Between 5 1584.97 316.99 s2b = (MSQb – MSQw)/n
sires = (316.99 – 147.77)/5 =
33.85
Within sires 24 3546.40 147.77 s2w = MSQw = 147.77
Total 29 5131.37

t = s2b/s2b + s2w = 33.85/(33.85 + 147.77) = 33.85/181.61

= 0.19

h2 = 4t = 4 x 0.19 = 0.76
 Sib analysis
• The estimation of heritability from half sibs needs
more detailed elaboration.
• A common form in which data are obtained is as
follows: a number of males (sires) are each mated
to several females (dams), the males and females
being randomly chosen and randomly mated.
• A number of offspring from each female are
measured to provide the data. The individuals are
measured from a population of half-sib and full-
sib families.
• An analysis of variance is then made.
• The phenotypic variance is divided into
components attributable to differences
between the progeny of different males
(the between-sire component, σ2S), to
differences between the progeny of
females mated to the same male (the
between-dam, within-sire component,
σ2D), and to differences between
individual offspring of the same female
(within-progeny component, σ2W).
• Suppose there are s sires, each mated to d
dams, which produced k offspring each.
The analysis will be as shown in the Table
below.
Source df Sum of Mean Composition of
square square mean square
Between s – 1 SSS MSS = σ2W+kσ2D+dkσ2S
sires
Between s(d-1) SSD MSD = σ2W+kσ2D
dams
(within
sires)
Within sd(k-1) SSW MSW = σ2 W
progenies
• The mean square within progenies is itself
the estimate of the within-progeny variance
component, σ2W.
• The between dam component is estimated
as σ2D = (MSD – MSW)/k
• The between-sire component is estimated as
σ2S = (MSS – MSD)/dk, where dk is the
number of offspring per sire.
• The estimate of the phenotypic variance is
give by: VP = σ2T = σ2S+ σ2D + σ2W .
• σ2S = ¼VA
• σ2D = ¼ VA+¼ VD+VEc
• σ2W = ½ VA+¾VD+VEw
Sib correlations Estimates of heritability
Half sibs: Sire-component
t(HS) = σ2S/σ2T h2 = 4σ2S/σ2T

Dam-component
h2 = 4σ2D/σ2T

Full sibs Sire + Dam:

t(FS) = σ2S +σ2D/ σ2T h2 = 2(σ2S+σ2D)/σ2T
 Example for 2 sires mated to 2 dams and each dam
produce 2 offspring

Source Mean Component of Heritability

square variance estimates
Between 6.03 σ2S = (6.03 – 3.81)/2x2 h2 = (4x
= 0.555 0.555)/3.895 =
sires 0.57

Between 3.81 σ2D = (3.81 – 2.87)/2 h2 =

= 0.47 (4x0.47)/3.895 =
dams within 0.48
sires
Within 2.87 σ2W = 2.87

progeny
• The choice of what sort of relatives to use for the
estimation of heritability depends on which types
of relatives are available.
• In addition to available relatives, precision and
bias should be considered.
• Generally, the closer the relationship, the more
precise is the estimate.
• Bias can be introduced by environmental sources
of covariance, and in the case of full sibs by
dominance.
• The half-sib correlation and the regression of
offspring on father are the most reliable estimates.
• The full-sib correlation is the least reliable due to
the component of common environmental and
dominance variance.
 Realized heritability
• Realized heritability is the h2 that is obtained from a
selection programme.
• Realized h2 is a function of the selection differential and
the response to selection.
• h2 = R/S
• Half-sib, parent-offspring and mid parent-offspring
analyses provide accurate estimate of h2. Full-sib analysis,
really does not provide h2, because it confounds VA with
VD. Realized h2 gives the percentage of VP that was
exploited in that generation.
• The methods are based on the assumption that the
phenotypic similarity between related individuals is caused
by the genes they have in common, and nothing else.
 Genetic correlation
• The phenotypes of two or more traits may be
associated or correlated, this means that the traits
do not vary independently.
• The extent to which performance in one trait is
associated with performance in the other trait is
called the phenotypic correlation (rP). This
correlation might be genetic or environmental in
origin (or a combination of the two).
• The phenotypic correlation between two traits can
be computed by measuring two phenotypes on a
number of individuals and then calculating a
correlation coefficient.
• One reason for a phenotypic correlation among
traits is that the traits are influenced by a common
set of genes.
• The extent to which two traits are determined by
the same set of genes is called genetic correlation
(rG).
• Genetic correlation is usually caused by pleiotropy
(one gene affects two or more traits).
• Genetic correlations may be positive or negative.
A positive correlation means that genes causing an
increase in the magnitude of one trait bring about
a simultaneous increase in the magnitude of the
other. Negative genetic correlation means that
genes that cause an increase in one trait produce a
corresponding decrease in another trait.
 Role of environment in
phenotypic expression
• Genes are the blueprints that are used to produce
the phenotypes.
• They produce these phenotypes in conjunction
with the environment.
• The environment influences the production of all
phenotypes, but quantitative phenotypes are more
affected by environmental factors than qualitative
phenotypes.
• Environmental factors that influence quantitative
phenotypes range from stocking density and feed
quality to female age, female size, spawning date
and feeding practices.
• Even if the environment plays a large role in the
production of a quantitative phenotype, the role it
plays is not critical to the success of a breeding
programme if it is the same for all fish.
• When conducting a selective breeding programme
to improve a quantitative phenotype, it is crucial
to be able to control environmental variables and
to prevent them from varying among individuals,
families and ponds.
• If the environmental variables are not controlled,
you will not know if the selected fish are best
because they are genetically superior or because
they had the better environment.
• Note that only fish that are superior genetically
will be able to transmit this superiority to their
offspring. This is the goal of selective breeding
programme.