Quantitative Genetics

Topic 1. Continuous and Discontinuous Variation

1 3 1 2 1
Above Left: Phenotypes are of two discrete forms.
Ratio of phenotypes is 3:1. Such a situation is seen
with 1 gene, two alleles, and complete dominance
and recessiveness
Above Right: Phenotypes are of three discrete
forms. Ratio of phenotypes is 1:2:1. Such a
situation is seen with 1 gene, two alleles, with the
heterozygote having some intermediate phenotype
If you considered three or four or more genes which control
some trait you might see additional discrete phenotypes. As the
number of genes increase, the curve becomes smoother and
smoother, i.e. continuous.

Birth weight of
babies born to
teenagers in
Portland, 1992

A lot of traits look like the graphs above…CONTINUOUS. This is due to

additive incremental effects of many genes. These genes and their
effects on phenotype are the subject of “quantitative genetics”.
Such curves also reflect the influence of environment, which tends to
smooth things out even more.
Polygene hypothesis

1909, wheat kernels

Observation: 6
Parental cross (true breeding)
4 4
red kernel X white kernel

F1 all pinkish kernels 1 1

F2 15:1 reddish to white

Result: See four shades of red

with distribution 1:4:6:4:1 from
more red to less red to white

Note that this is an “out of 16” set

that, like earlier examples, differs
from 9:3:3:1
Explanation:
Two genes, R and C
Both R and C contribute to redness

P RRCC X rrcc
full red white
4X 0X
F1 RrCc
pink
2X

F2 range of 16 genotypes
RRCC to rrcc
4X 0X

Color is determined by how many R

and C are present in order of
RRCC > RRCc=RRcC=RrCC=rRCC >
RRcc>etc….down to rrcc
(the wheat genome is composed of a
hybrid of three similar genomes. This
accounts for the presence of multiple Above picture uses genes
genes that all contribute to kernel redness) labeled R and R’
Well actually, the truth is, it’s a three gene system……
 Topic 2. Normal Curves and the Statistics of Variation

To examine the range

of variation in a
population you
collect a sample and
hope it is
representative of the
larger population.

From the
measurements you
can plot a
FREQUENCY
DISTRIBUTION.

Such data is often

modeled by a
NORMAL curve.
Statistical Measures that Describe Normal Curves
…These are not that hard…

Average (mean).

x(avg) = sum of all individual x measurements divided by

number of measurements

Variance.

variance = s2 = E(xi-x(avg))2 / n-1

note that variance is a squared term. Variance gives you a

measure of how spread out the curve is

Standard Deviation = √s2

Standard deviation is another measure to describe how spread

out the curve is. The nice thing is that the units are the same
as the original measurements (i.e., not squared)
Examples of
normal curve
distributions that
have identical
phenotypic
averages (at 10)
but different
extents of
variation away
from the average,
here measured as
variance.
A further description of
a normal curve is
shown. Here the curve
is described by its mean
(average) and by
distance in standard
deviation away from the
mean (+1s, -1s; +2s, -2s,
etc).
The percentage values
at the top show the
proportion of individuals
in a population that are
expected to fall within a
given standard
deviation.
So, 99% of the
population lies within +/-
3s.
Book example showing
calculations of mean,
variance, and standard
deviation.

Why use these

statistical approaches?
Because you need
ways to describe the
way traits vary
continuously. Such
traits involve multiple
genes (alleles) adding
up to generate
phenotypes.
 Covariation and Correlation

Sometimes it is interesting to ask if two measurements (e.g.

phenotypic traits) are related or correlate with one another.
In other words, as the measurement of trait x varies, does the
measurement y also vary with it? That is to say, might there be a
connection between x and y?

So of course there are formulas for that sort of thing…

Covariance
covariancexy = E(xi-xavg)(yi-yavg) / n-1

Correlation coefficient
correlation coefficient r = covxy / sxsy
where s is standard deviation for either x or y
The correlation coefficient r provides a measure of how
closely two observations (x, y) are related….

r ranges from -1 to +1

r = +1 is a perfect positive correlation such that when x changes

y changes accordingly, making perfect line
x goes up, y goes up

r = -1 is a perfect negative correlation such that when x

changes y changes accordingly in the opposite
direction making perfect line
x goes up, y goes down

r=0 means no correlation, the values of the x and y

measurements are unrelated. For whatever x is, y can
be anything

No cause or effect is implied by correlation values. They only show how

two variables change with respect to one another but they do not say that
x causes y or y causes x or that both are due to the same underlying
biological mechanism, or not.
Examples of what data
corresponding to
correlation values 0, 0.3,
0.5, 0.7, -0.7, and 0.9 might
look like (so says the book)

Data:
Father height vs. Son height
Next: A few examples of quantitative
genetics…
 Genetics
Example
True breeding
short corn X
True breeding
long corn
Both have
distinct
distributions
around some
average.
Offspring have
sizes that are
intermediate.

Offspring are heterozygotes since parents were true breeding. Alleles

causing small size have combined with alleles causing large size, resulting in
intermediate size. Variation is environmental. F1 X F1 generates F2 with
mean similar to F1. But, notice, there is much greater variance, extending as
far out as the average of the original parents. Since it is unlikely that the
environment affected F2 more than F1, we conclude that the greater variation
is due to assortment of alleles into new combinations.
Concept: Transgressive segregation
In the previous case breeding resulted in the restoration of traits near
the original parents (long ear and short ear). Here we ask if a
population of offspring from two parental populations could extend
past the variation seen with the original parents.

Genetics Example
F2 are more extreme than the grandparents!
How can this be?
Parental
Let one parent be aaBBcc
Let the other parent be AAbbCC

Assign value of 5 to each capital letter

F1 So parent 1 is 2 X 5 = 10
And parent 2 is 4 X 5 = 20
F1 offspring are AaBbCc, making them
3X5=15
F2
But cross of F1 generates F2 offspring with
ranges from aabbcc (0) to AABBCC (30)
 The relation between the seed weight of the parental generation and the filial Johannsen
generation in a variety of brown beans. Experiment
The figures in the table represent the filial numbers of beans of the different weight
categories Select beans
for weight.
weight of the filial beans
weight of the weight categories Self cross
parental beans
10 20 30 40 50 60 70 80 90 average Measure
20 - 1 15 90 63 11 - - - 43,8 bean weight
30 - 15 85 322 310 91 2 - - 44,5 produced by
40 5 17 175 776 956 283 24 3 - 44,2 offspring.
50 - 4 57 305 521 196 51 4 - 48,9
60 - 1 23 130 230 168 46 15 2 51,9
70 - - 5 53 175 180 64 15 2 56,0

total 5 38 170 1676 2255 928 187 33 2 47,92

according to W. JOHANNSEN; 1903, 1926

Bean weight of offspring are similar to parents, although a LOT of variation

Bean weight in offspring of extreme parents tend to go to the average. This is
called regression to the mean. Organisms at extreme ends of the curve may
result from (unusual) environmental influences and may not be inheritable.
 Topic 3. Genes and the Environment

Phenotypic variance (Vp) (range of color, range of size, etc) of a

trait in a population should have two main contributing
components…

Vg genetic component (variation due to genes)

Ve environmental component (variation due to environment)

In its most simple form this gives a kind of crude equation…

Vp = Vg + Ve

Which really just says that variation in how things are depends
on
1) genes
2) environment
Examples of the
effect of
temperature
(environment) on
two related plant
genotypes.
The Book has an Interesting Example about Genes/Environment:

Suppose you want to measure the variance in mouse weights and determine
how much is due to environment and how much is due to genetics.

Case 1. Population of inbred mice homozygous at every allele.

The phenotypic variance Vp = 0.3 square ounces for these inbred mice

Note: recall that variance is squared.

Note: The value 0.3 is measured experimentally.
We put this value into the ‘equation’

Vp = Vg + Ve to give 0.3 = 0 + Ve and Ve = 0.3 = Vp

Why is Vg = 0 here? Because the mice are inbred. They are genetically
identical to one another, same alleles of every gene are present in every
mouse, they really are ‘clones’ of one another. So, there is no contribution
from genes to control how different mice vary in size.

Therefore, the phenotypic variance is due to environmental effects only.

Case 2. Population of wild (outbred) mice composed of individuals with
many different allele combinations.

Genetically these mice are very diverse!

Vp = 0.54 square ounces for the outbred mice.

Note: As before, this value is measured experimentally. The mice are

put on the scale.

If the mice are grown under the same conditions as for the inbred mice
(same environment), then the variation due to environment should be
the same as in the first example. We calculated that in case one and
found that Ve = 0.3.

Assuming that, then

Vg (variation due to genetic component) = 0.54 – 0.3 = 0.24 square

ounces. So we crudely partition a range of phenotypic measurements
into environmental and genetic components.
(The book extends this discussion to include dominance, epistasis, etc but I am
not covering that)
 Topic 4. Heritability

Total Phenotypic Variation = VG + VE

Question is…how much of one or the other?? Nature vs. Nurture

The concept of heritability is a way of assessing how much of the

phenotypic variation in a population is due to a genetic component.

Understanding heritability is important for things like…

--making predictions about the phenotype of a population of offspring

based on the phenotypes of a population of parents

--estimating whether artificial selection (breeding for particular qualities or

traits) is likely to be successful.

Heritability is divided into two main types:

Broad Sense and Narrow Sense

‘Broad sense heritability’ (H, or Hb) tries to evaluate how much
phenotypic variation is due to genetics. It is supposedly a
squared term to reflect the fact that variances are squared.

It is denoted as the fraction or proportion of phenotypic variation

that can be attributed to genes (Vg) compared to total amount of
phenotypic variation (Vp)

Broad Sense Heritability = Vg / Vp

‘Narrow sense heritability’ (h, or Hn) tries to quantify how much
phenotypic variation is due to genes that ‘add up’ and contribute to the
phenotype (additive genetic variation).

Narrow sense heritability considers genes in an organism that may

have small, incremental additive effects on a trait. It wishes to exclude
things like dominant/recessive and epistatic type effects.

This turns out to be the more useful number because it helps predict
whether one will be successful in breeding animals to produce
improved traits.

Mathematically, narrow sense heritability (Hn) is the fraction or

percentage of phenotypic variation that can be attributed to additive
genetic effects compared to the total phenotypic variation for that trait.

Narrow Sense Heritability = Va / Vp

Some Examples of Heritability Values
--Heritability is complicated and the book has a long section on it which is
hard to follow. It emphasizes that heritability is calculated for a given time
and place. Please keep in mind that our goal is to achieve a basic conceptual
understanding and to use these ideas in relatively simple problems.

--Additional examples describe inheritance of things like number of eyes

(heritable but zero variance), as well as effects of diet and environment.
 Topic 5. Artificial Selection

If an organism were reduced to numbers, it would be a

collection of measurements that represent the millions or
billions of possible measurable traits. All those data points
collectively will determine how prolific that particular organism
is in producing new offspring. The idea of natural selection is
that the data points shift from generation to generation due to
the fact that some traits (i.e. alleles) confer greater fitness (more
reproduction) than others. Alleles that contribute to successful
reproduction tend to increase in frequency as time passes.
That is, some alleles become common, others decline.

In artificial selection we try to move left or right along some

phenotypic measurement curve by selective interbreeding of
animals with a desired trait. The purpose is usually for making
food production more efficient, but sometimes its just for
decoration.
Genetics Example:
We select fruit flies for avoidance
(negative phototaxis) or attraction
(positive phototaxis) to light.

Selection works for a while then it slows

down and levels off (selection limit).
Why?

1. No more genetic variation controlling

phototaxis was present; in other words,
the ability of genes to control
phototaxis was maxed out. The best
possible combination of alleles that
contribute to phototaxis was obtained.
The flies had been inbred to maximum
extremes of phototaxis.

2. Further selection for positive or

negative phototaxis was associated
with severe problems so that the flies
become dysfunctional.
The success of selection can be related to narrow sense heritability
and on the selection differential.

R = Hn2 x s

R is selection response (how good selection has worked in the offspring)

s is selection differential (the difference between selected individuals at the start

compared to the average)

Hn2 is narrow sense heritability (a measure of how much of the trait is under the
control of (additive) genetic variation).

That is, success in an artificial selection experiment depends on

what your chances were (how much is genetic) and where you
started from along the phenotypic distribution curve.

Another way to put this is to rearrange so that Hn2 = R/s

Or……
Narr. sense herit. = (F1 mean – original mean)/(P mean – original mean)
The Book Example:

Suppose we have a population of fruit flies: Mean body size is 1.3 mg

Half the population breeds together to produce offspring. The F1 flies also have
a 1.3 mg mean body size, same as the parental population.

The other half of the population is used to select out flies with mean body size of
3.0 mg. These flies are intebred to generate an F1. The F1 flies have a mean
body size of 2.0 mg.

The “selection response”, or R, here is 2.0 mg – 1.3 mg = 0.7 mg

0.7 mg = Hn2 s

The “selection differential” is 3.0 – 1.3 = 1.7 mg

0.7 = Hn2 X 1.7

so, 0.7 / 1.7 = Hn2

0.41 = Hn2 (h2, h, narrow sense heritability)

Artificial
selection in
dogs has
produced a
wide range of
phenotypes.

In addition,
gene mutations
may
contribute.
People
wonder what
genes are
involved in
how dogs
look, how
many genes,
what those
genes do.
Size
Fur quality
Color
What is a feral
dog? Why does
a feral dog look
like a feral dog?
Artificial
selection
and
appearance
of novel
mutations.
 Although this individual wouldn’t run very far if a
lion was after it, it did make its way into a painting
 Topic 6. Correlations between two traits…

…may occur if genes that have an effect on phenotype x also have an

effect on phenotype y (aka pleiotropic effects, genes involved affect more
than one thing at once, i.e. trait 1 and trait 2)

…may occur because of linkage. That is, a gene that controls phenotype x
is very close (linked) to a gene that controls phenotype y. This means
alleles at both genes tend to get inherited together (i.e. meiotic
recombination does not often occur between the genes so that they
become separated). We will talk more about this later.

…are important in thinking about breeding and artificial selection

experiments. Sometimes a breeder wants to select for a particular trait,
but there are negative side effects such that improvement of X is
associated with decline in Y. Which means that organisms can’t be bred
to get faster/smarter/bigger etc to some infinite point because there is
usually some (hidden) disabling cost or tradeoff.
Some examples of genetic correlation values
Statistical and Genetic Concepts in this Chapter

This is not a statistics course, and so our application of these concepts is

fairly simple. My concern is that you can follow the lecture material and
understand it for concepts and practical applications, and that you can
calculate some basic statistical and genetic measures.

Use the book problems as a guide, but bear in mind that some of them are
pretty tough. The topic of quantitative genetics is never presented very well
as a single chapter like this, it comes out a bit fuzzy, or very advanced, or
both. Sorry.

We will not make lines based on regression analysis, and we are not going to
cover the section on quantitative trait loci (QTLs)

Aside from concepts in lecture notes, you should be familiar with average
(mean), variance, standard deviation, correlation, normal curves,
broad/narrow sense heritability, relationships such as…

Vp = Vg + Ve, Hb2 = Vg / Vp, Hn2 = Va / Vp R = Hn2 s

Hn2 = (next gen – orig) / (selected – orig)

The moral of the story…
Quantitative genetics is important because many genetic
phenomena can only be described in terms of the contribution of
many (5, 10, 100, 1000) genes, as opposed to simple ‘up-down’
dominance-recessive relationships or ‘one gene affects another’
epistatic relationships. Individuals are the product of many genes
acting together during development. Its not like there is a nose
gene, an eye gene, an elbow gene, but rather many alleles
collaborate to generate complex outcomes. Quantitative genetics
is more difficult than Mendelian genetics. But it is of interest
because it starts to address how all our genes interact to produce
distinct individuals and complex diseases.

One often come across news and public debate about all kinds of
issues such as heritability of IQ, predisposition to cancer,
similarities between monozygotic and dizygotic twins, ability of
the agricultural industry to select for particular traits in plants and
animals, basis of complex diseases such as alcoholism, diabetes,
asthma, neurodegeneration, autism, etc. Pay attention to this kind
of news, a lot of it concerns quantitative genetics.
Commercial Production
of Chickens Takes Toll on
Genetic Diversity
By HENRY FOUNTAIN
Published: November 3,
2008
To the connoisseur of fine food, chicken may seem depressingly monotonous no
matter how it’s prepared. But scientists worry about a more basic degree of
sameness — a lack of genetic diversity in the birds that are raised for meat and
eggs. An analysis of commercial chicken populations around the world by William M.
Muir of Purdue University and colleagues has revealed the extent of the problem.
Fifty percent or more of the diversity of ancestral breeds has been lost, they report in
The Proceedings of the National Academy of Sciences. That could make chicken
production more susceptible to disease outbreaks for which resistant genes have
disappeared.
Sampling about 2,500 birds, the researchers looked at several thousand instances of
genetic variation and used that to estimate what a hypothetical ancestral population
looked like genetically. “Then we were able to say what is missing” in commercial
birds, Dr. Muir said.
Their findings indicate that most of the diversity was lost with the advent of wide-
scale commercial production in the 1950s. Only a handful of hundreds of breeds
have been crossed to produce broilers and layers.