Quantitative Genetics
The genetic basis of many traits is only poorly known. We lack specific information about:
(Remember that even the two-locus model is not completely understood!)
Evolutionary models fall into two main camps:
Phenotype: The visible [or measurable] properties of an organism that are produced by the interaction of the genotype and the environment.
In quantitative genetics, the phenotypic value (= P) of an individual (e.g. height) is attributed to the genotype of the individual and to its environment:
The genotypic value (= G) is a measure of the influence of every gene carried by the individual on the phenotypic value.
The environmental deviation (= E) is a measure of the influence of the environment of an individual, scaled such that the average value of E is zero within a population.
An additional component is possible when the environment and genotype interact to influence the phenotype of an individual. This is called genotype-by-environment interaction (GxE). (We will ignore this complication.)
Example: (From Intermediate Genetics Course, by P. McClean at North Dakota State Univ).
Genetically uniform strains of wheat were grown in Casselton, North Dakota by Dr. Jim Anderson.
| Yield (bushels/acre) | |||
|---|---|---|---|
| Genotype | |||
| Year | Roughrider | Seward | Agassiz |
| 1986 | 47.9 | 55.9 | 47.5 |
| 1987 | 63.8 | 72.5 | 59.5 |
| 1988 | 23.1 | 25.7 | 28.4 |
| 1989 | 61.6 | 66.5 | 60.5 |
| 1990 | 0.0 | 0.0 | 0.0 |
| 1991 | 60.3 | 71.0 | 55.4 |
| 1992 | 46.6 | 49.0 | 41.5 |
| 1993 | 58.2 | 62.9 | 48.8 |
| 1994 | 41.7 | 53.2 | 39.8 |
| 1995 | 53.1 | 65.1 | 53.5 |
For the roughrider strain of wheat, the average yield (= the phenotype) over the ten years was 45.6 bushels/acre. This would be an estimate of the genetic (G) component of roughrider's phenotype.
In 1986, the environmental deviation was +2.3 for roughrider:
Consider one locus with two alleles (A1 and A2):
Here we have arbitrarily assigned a scale to the genotypic values such that A1A1 individuals are given the value +a and A2A2 individuals are given the value -a.
(This notation follows that standardly used in the quantitative genetics literature.)
The mean phenotypic value at any point in time will depend on the allele frequencies:
= p2 (a) + 2pq (d) + q2 (-a)
[Remember that E is measured such that mean(E)=0.]
If a parent reproduced clonally (like Dolly), then its offspring would inherit the complete genome of the parent. In this case, the parent and offspring would have the exact same value of G.
If a parent reproduces sexually, however, only one allele at each locus will be passed on to the offspring.
The next step is to measure the average effect on the phenotype of this one allele:
Average effect of an allele: The mean phenotype of individuals which received that allele from one parent when the other allele is chosen at random from the population, measured as a deviation from the population mean.
Since a randomly chosen allele will be A1 with probability p and will be A2 with probability q, the average effect of allele A1 would be
If a parent is homozygous at a locus, it cannot transmit this status to its children.
Only one allele is passed to the offspring, so whether the offspring will be homozygous or not depends on the allele frequency within the rest of the population.
The average effect measures how offspring that inherit a specific allele differ from the population as a whole.
The advantage of defining the average effect of an allele transmitted from parent to offspring is that one can sum these effects up across many loci to estimate the average influence that a parent will have on their offspring through genetic inheritance!
(WARNING: This ignores interactions between loci.)
Breeding value of an individual (A): The sum of the average effects of all alleles carried by an individual.
Since only half of these alleles are transmitted from a parent to its offspring, the phenotype of offspring will, on average, be half the breeding value of the parent.
This provides a second definition of the breeding value: as twice the difference between the mean of the offspring of an individual and the mean of the population.
MEASURABLE
If you take the genotypic value of an individual (G) and subtract off its breeding value (A), you are left with terms that measure interactions among the alleles that the individual happens to carry.
These interactions are of two sorts:
= A + D + I + E,
but it is primarily the breeding value (A) that influences the phenotype of the offspring of this individual.
Example: To increase milk yield, dairy farmers estimate the breeding value of bulls from the average dairy production of the daughters of each bull. Those bulls with a higher breeding value are then used to produce the next generation of cows.
Say that the daughters of a particular bull (mated to several cows) produce 100 liters of milk per day in a herd with an average production of 75 liters. The breeding value of the bull will then be estimated at 125.
Say that a particular cow produces 100 liters of milk per day compared to an average of 75. Her daughters (when mated with different bulls) produce 80 liters of milk per day. The mother cow is estimated to have a breeding value of 85.
The difference between her actual yield (100) and her breeding value (85) will be due to rearing environment and/or interactions among alleles (at the same or different loci).
Quantitative genetics is particularly concerned with describing the variation within a population and with estimating the genetic component of this variation.
For many traits, phenotypic values vary greatly within a population (from Christiansen and Feldman, 1986):
The variance of phenotypic values (VP) is an important measure in quantitative genetics.
Just as the phenotype of an individual may be due to genetic and/or environmental influences, the variability within a population may be due to genetic or environmental differences among individuals.
The phenotypic variance is sub-divided into genotypic variance (VG) and environmental variance (VE):
(We are ignoring any correlation or interaction between genes and environment.)
The genotypic variance describes how much phenotypic variance is due to genetic differences within a population.
Let's now ask how similar, on average, are parents to their offspring?
But just as the genotypic value of an individual is not passed in its entirety to offspring, only a component of the genotypic variance explains the resemblance of relatives within a population.
The genotypic variance is further sub-divided into:
The additive genetic variance (VA) equals the variance in breeding values of individuals within a population.
Example: The dairy yields of 10 cows are: 75, 88, 52, 83, 82, 43, 100, 48, 79, 100.
The phenotypic mean is 75.
The phenotypic variance is estimated as 425.6, calculated from:
The breeding values of these same cows are: 74, 88, 56, 83, 82, 57, 93, 54, 84, 81.
The mean breeding value is 75.2.
The additive genetic variance is estimated as 205.5 (also calculated from the above formula).
These quantities can be used to determine how much relatives will resemble one another.
They also determine how much evolutionary change will be accomplished when certain parents reproduce while others do not.
Broad Sense Heritability (H2): The extent to which variation in phenotype is caused by variation in genotype (= VG/VP).
Broad sense heritability will be 1 if all of the phenotypic variation within a population is due to genotypic differences among individuals.
Broad sense heritability will be 0 if all of the phenotypic variation is caused by environmental differences.
Example: The roughrider strain of wheat is genetically uniform. Therefore, all variation in yield among plants is due to environmental differences among plants. H2 = 0.
Heritability is measured in a particular population at a particular place and time.
Observing H2 = 0 in roughrider does not mean that wheat yield would have no genetic component in other populations of wheat.
Narrow Sense Heritability (h2): The extent to which variation in phenotype is caused by genes transmitted from parents (= VA/VP).
Narrow sense heritability will be 1 only if there is no variation due to genetic interactions (dominance or epistasis) or to environmental differences. When h2 = 1, P = G = A.
Narrow sense heritability can be zero even if broad sense heritability is not (BUT NOT VICE VERSA) because all the genetic variation within a population may be due to dominance or epistasis.
Example: For the ten cows, VP = 425.6 and VA = 205.5. The narrow sense heritability is 205.5/425.6 = 0.48.
Example: Variation in human birth weight (from Penrose, 1954; Robson, 1955; see Falconer, 1989)
| Cause of variation | % of total | |
| Genetic | 18 | |
Additive | 15 | |
Non-Additive | 1 | |
Sex | 2 | |
| Environmental | 82 | |
Maternal genotype | 20 | |
Maternal environment | 24 | |
Age of mother | 1 | |
Parity | 7 | |
Intangible | 30 |
A very large portion of the phenotypic variability is environmental in origin.
The broad sense heritability of birth weight is only 0.18 and the narrow sense heritability is 0.15.
(We'll assume that mating is random and that epistasis and gene-by-environment interaction can be ignored.)
A parent with genotypic value G will have an offspring whose genotypic value will, on average, be A/2 (half of the parent's breeding value).
It can be shown (see proof if interested) that the covariance between the phenotype of a parent and its offspring will therefore equal 1/2 VA.
More importantly, the correlation between the phenotypic values of parents and offspring will equal 1/2 VA/VP = 1/2 h2.
[Assuming that the phenotypic variance does not change in the parental and offspring generations.]
One can therefore use the correlation between offspring and parents to estimate the narrow sense heritability:
If both parents are measured, the regression of offspring values on mid-parent values is:
The correlation between two full-siblings is:
The expected correlation between many other types of relatives has also been calculated.
Sample heritabilities (from Falconer, 1989):
| h2 | ||
| Man | ||
| Stature | 0.65 | |
| Serum immunoglobulin level | 0.45 | |
| Cattle | ||
| Adult body weight | 0.65 | |
| % butterfat | 0.40 | |
| Milk-yield | 0.35 | |
| Pigs | ||
| Back-fat thickness | 0.70 | |
| Weight gain per day | 0.40 | |
| Litter size | 0.05 |
Narrow-sense heritability is particularly important in animal and plant breeding, because it predicts what the will happen in the next generation when particular parents are chosen to breed.
Selection differential (S): The phenotypic mean of parents chosen to breed minus the population mean.
Response to selection (R): The phenotypic mean of offspring of these parents minus the population mean.
The most important formula in quantitative genetics states that:
Characters with a high heritability (e.g. back-fat thickness in pigs) will respond rapidly to selection, whereas characters with low heritability (e.g. litter size in pigs) will respond slowly.
EXAMPLE: Clayton, Morris, and Robertson (1957) selected on abdominal bristle number in Drosophila melanogaster.
A previous estimate for heritability of bristle number within this population was 0.52.
The parental mean was 35.3 bristles.
Among those allowed to breed (those selected), the mean number of bristles was 40.6.
The selection coefficient was S = 40.6-35.3 = 5.3.
The response to selection was expected to be R(exp) = 0.52*5.3 = 2.8.
The offspring generation actually had 37.9 bristles on average, for an observed response of R(obs) = 37.9 - 35.3 = 2.6.
(Reasonably close!)
Heritability is not a constant attribute of a population.
As the population evolves, the heritability of a trait will change as:
Often, however, heritability remains roughly the same over a number of generations.
This can be seen if we plot response against the cumulative selection differential; if heritability remains constant, this gives a straight line.
EXAMPLE: Six-week weight in mice selected over ten generations (Falconer, 1973)
Here h2 is estimated to be 0.398 in the case of upwards selection and 0.328 in the case of downwards selection.
After longer periods of time, the response to selection on a quantitative trait often reaches a plateau. For example (from Falconer, 1989, based on Roberts, 1966):
The total response before a plateau is reached depends on many factors.
(1) The total response will be less when few individuals are chosen to breed, since less genetic variation is preserved among these individuals.
(2) The total response will be less when selection occurs rapidly because of genetic hitchhiking (some alleles that act in the opposite direction may get dragged along and fix, especially when S is high).
(3) The total response will be less if few loci contribute to the trait, since those few loci will go to fixation and since the array of possible combinations of alleles is much more limited.
In practise, animal and plant breeders have to juggle a number of factors (financial, logistic, biological) in an attempt to optimize the response of a bred population to selection.
SOURCES:
Back to Biology 336 home page.