You are on page 1of 45

Introduction to Quantitative Genetics

Nicolas Salamin
Department of Computational Biology

nicolas.salamin@unil.ch; tel: 4154; office: Génopode 2025.3

March, 2024

Link to the Moodle course: click here


Teams channel: click here
What is quantitative genetics?

from www.globalbiofest.com

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 2


What is quantitative genetics?

Key topic in evolutionary biology


Theory of the statistical relationship between genotypic variation and
phenotypic variation

Main questions
1. what is the cause of phenotypic variation in natural populations?
2. what is the genetic architecture and molecular basis of phenotypic
variation in natural populations?

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 3


Level of study of traits

Patrick Phillips, NiMBIOS lecture 2016

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 4


A bit of history...

After the rediscovery of Mendel’s work, heated debate between two


“schools”:

Mendelians Biometricians
▶ discrete traits inherited and ▶ supporters of Darwin and
associated to genotypes founders of modern statistics
▶ evolution moves by major leaps ▶ important variation in evolution
due to new mutations did not follow Mendel’s rule
▶ merging of phenotype and ▶ e.g. height, weight, BMI, skin
color
genotype in continuous traits
▶ vary continuously
▶ if affected by environment, then ▶ blending inheritance
not inherited ▶ respond to environment

Modern synthesis: fusion of Darwin’s evolution by natural selection with


Mendelian inheritance.

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 5


From Mendel and Pearson to now

Patrick Phillips, NiMBIOS lecture 2016

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 6


Genotype vs phenotype

Genotype
▶ genetic makeup of an organism
▶ specific set of alleles inherited at a locus

Phenotype
▶ any measurable characteristic of an individual, e.g. height, hair color,
disease status
▶ selection acts on the phenotype
▶ modified by genetic and environmental background
▶ monogenic traits (rare)
▶ discrete binary characters
▶ polygenic traits (common)
▶ discrete (number of bristle), continuous (body mass) or binary (disease
status)

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 7


Increasing the number of loci

leads to a continuous phenotype frequency


▶ with 1 locus → 2 homozygous, 1 heterozygous
▶ to many loci → continuous distribution

cubocube.com

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 8


Quantitative genetics model

Writing down the model for 1 locus


( AA = +0.95 )
phenotype =µ+ Aa = +0.99 +E
aa = +0.76

to multiple loci
( AA = +0.95 ) ( BB = +0.01 ) ( CC = −0.60 )
phenotype = µ+ Aa = +0.99 + Bb = −0.05 + Cc = +0.35 +E
aa = +0.76 bb = −0.02 cc = +0.06

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 9


Phenotype distribution

Effect of many loci


▶ overall mean = 0

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 10


Phenotype distribution

Effect of many loci


▶ overall mean = 0
▶ locus mean = 0.90
▶ all others vary around
overall mean, but
effects will add up

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 11


Genotype effects on phenotype

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 12


Reaction norm

Distribution of phenotypes
▶ distribution of environments reflected as a distribution of phenotypes
▶ transformation of environmental distribution into phenotype
distribution is determined by the norm of reaction

from book: Griffiths et al 1996 Freeman and Company

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 13


Galton’s regression to the mean

https://select-statistics.co.uk/blog/regression-to-the-mean-as-relevant-today-as-it-was-in-the-1900s/

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 14


How to obtain it? Try and explain

k<-100; n<-1000 #k=num loci, n=num parent/offspring


p<-0.5 #allele frequency

d<-data.frame(midPar=array(dim=n),offVal=array(dim=n), dad=array(dim=n), mum=array(dim

for(i in 1:n) {
#genotypes
dad<-rbinom(k,2,p); mum<-rbinom(k,2,p)
kid<-rbinom(k,1,dad/2) + rbinom(k,1,mum/2)

#phenotypes / sd=5 is arbitrary


z_dad<-rnorm(1,sum(dad),5); z_mum<-rnorm(1,sum(mum),5)
z_kid<-rnorm(1,sum(kid),5)

d$dad[i]<-z_dad; d$mum[i]<-z_mum
d$midPar[i]<-mean(c(z_dad,z_mum))
d$offVal[i]<-z_kid
}

plot(d$midPar, d$offVal, xlab="Mid-Parent trait",


ylab="Child trait")

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 15


The wrong blending inheritance

Half of variation removed in each generation


Blending inheritance of continuous traits

σ 2 (zf ) σ 2 (zm ) σ 2 (z)


 
2 2 zf + zm
σ (zo ) = σ = + =
2 22 22 2
with zo is the offspring value, zf and zm are the maternal and paternal
values and trait variance is the same in females and males
(i.e. σ 2 (zm ) + σ 2 (zf ) = 2 × σ 2 (z)). It leads over time t to
 t
1
σ 2 (zt ) = σ 2 (z0 )
2

Mendelian inheritance
Transmission of alternative alleles in gametes of heterozygous parents
restores genetic variance → solve the issue and is integrated in the modern
synthesis

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 16


Partitioning phenotypic value

Basic model to use

P =G +E
where
▶ P is the phenotypic value measured on individuals
▶ G is the mean phenotypic value of those carrying the same genotype
▶ E is the deviation due to the environment, assumed with mean = 0

G and E are artificial constructs that need to be estimated using statistical


methods

Genotypic values are functions of the environments experiences (and we left


aside the G × E interactions)

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 17


Single locus influence

How does it influence a phenotype


▶ G is expected phenotype contribution (for a given set of genotypes)
resulting from the joint expression of all of the genes underlying the
trait
▶ for multilocus trait, G can be very complex
▶ what if a single autosomal locus affect the phenotype? (weird, but
easier to show the process and hold for many loci)

Genotypic value

 2a if genotype is AA
Genotypic value = (1 + k)a if genotype is Aa
0 if genotype is aa

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 18


Dominance and additivity

k and a values

 2a if genotype is AA
Genotypic value = (1 + k)a if genotype is Aa
0 if genotype is aa

▶ a is the additive genotypic value


▶ 2a = difference in mean phenotype for AA and aa
▶ k is the scaled measure of dominance
▶ k = 0 → alleles A and a are additive
▶ k > 0 → A has dominance over a (reverse for k < 0)
▶ k = 1 → complete dominance
▶ k > 1 → overdominance since mean phenotype of heterozygotes
exceeds both homozygotes
▶ k < −1 → underdominance

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 19


Example

Fecundity in Merino sheep


▶ litter size is polygenic, but mainly driven by Booroola gene in Merino
sheep (Piper and Bindon, 1988)
▶ for three genotypes, mean litter size for 685 records

Genotype Mean litter size


bb 1.48 (sd = 0.09)
Bb 2.17 (sd = 0.08)
BB 2.66 (sd = 0.09)

Small exercice (send me the answers)


▶ what is the dominance value k for this locus?
▶ is there significant dominance effect?

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 20


Other way to represent

Change the relative value


Genotype value relative to the average mean genotype for the two
homozygotes

 a if genotype is AA
Genotypic value = d if genotype is Aa
−a if genotype is aa

Answer the following question


▶ what is the value of d to cover all cases shown above (i.e. additivity,
dominance, . . .)?
▶ tip: add a to all scales and let d = ak
▶ k = d/a, which is why it is called the scaled measure of dominance

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 21


Representing genotypic values

Genotype aa Aa AA

Genotypic value −a d a
0

total difference: 2a

An example to illustrate
Calculate, with the following example
▶ the dominance effect d
▶ the additive genotypic value a

aa Aa AA
weight (g) 6 12 14

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 22


Decomposing genotypic values

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 23


Decomposing the genotypic values

G is a complex function across loci


Decompose G into additive and non-additive components

Gij = GijA + δij = µG + αi + αj + δij

for individuals with alleles i and j at a locus.

The function has the following components


▶ Gij is the mean genotypic value
▶ µG is the genotypic mean in the population (a fonction of P)
▶ αi and αj are the mean additive effects for each allele
▶ µG + αi + αj is the predicted genotypic value over all genotypes of
allele i and j
▶ δij is a dominance effect, i.e. difference between true and predicted
value from sum of average effects

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 24


Average effect of each allele

Decomposing the genotypic effect


Again, assume a trait locus with two alleles i = A and j = a. We then have

 µG + 2αi if genotype is AA
GijA = µG + αi + αj if genotype is Aa
µG + 2αj if genotype is aa

The goal is to calculate the average effect of each allele for a phenotype
▶ genotype values relative to the average of the mean genotypic value for
the two homozygotes
▶ help the interpretation of the additive effects for the alleles
▶ assume that A increases the phenotypic value (see graph in previous
slide)

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 25


Mean genotypic value

Assuming Hardy-Weinberg equilibrium


▶ frequency of A and a are p and q
▶ what is the mean genotypic value µG on this scale?

µG = a(p 2 ) + d(2pq) − aq 2
= a(p 2 − q 2 ) + d(2pq)
= a(p + q)(p − q) + d(2pq)
= a(p − q) + d(2pq)

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 26


Population mean genotype value

Contribution of a locus to the population mean


▶ two terms are defining it
▶ a(p − q), which only depends on the homozygotes
▶ d(2pq), which only depends on the heterozygotes
▶ under no dominance, mean genotypic value proportional to allele
frequency µG = a(1 − 2q) (or a(2p − 1))
▶ with complete dominance of the A allele (d = a, or d = −a for
dominance of a), mean proportional to the square of the allele
frequency µG = a(1 − 2q 2 )

Questions
▶ total range of values attributed to this locus is 2a
▶ what would be the contribution of this locus to the phenotype if the
population is monomorphic A? What if monomorphic a?

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 27


Mean allelic effect

Different from mean genotypic effect


▶ average effect linked with an allele carried by an individual (i.e. αi )
▶ important to characterize because parents transmit one of their alleles
at a locus to their offsprings, not their genotypes
▶ breeding value is the sum of the additive effects of the alleles carried by
an individual and that will be given to its offsprings


 µG + 2αi if genotype is AA
GijA = µ + αi + αj if genotype is Aa
 G
µG + 2αj if genotype is aa

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 28


Mean allelic effect, cont.

How to estimate αi ?
▶ αi is the mean effect of allele i = A
▶ mean deviation from population of individuals with allele A when the
other allele is drawn at random from the population
▶ assume as always that A and a have frequencies p and q
▶ genotypic value of AA is a, Aa is d and aa is −a

What is the average effect of the genotypes that are created if A is


randomly paired with gametes in the population?

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 29


Mean allelic effect, cont.

Estimation of αi
a × P(A pairs with another A) + d × P(A pairs with a a)
Then, average effects of genotypes with A is ap + qd

It follows that
αi = pa + qd − µG = q[a + d(q − p)]

Estimation of αj
Similarly,
αj = −qa + pd − µG = −p[a + d(q − p)]

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 30


Average effect of allele substitution

When we have two alleles


▶ suppose that we could change at random a alleles into A
▶ what would be the resulting change in genotypic value?
▶ mean change produced is the average effect of allele subsitution

How would this work out if a alleles are chosen at random


▶ proportion p are in Aa genotypes, which change into AA
▶ genoytpic value change from d to a → with an effect of a − d

▶ proportion q are in aa genotypes, which change into Aa


▶ genotypic value change from −a to d → with d + a effect

What is the average allele effect of allele substitution?

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 31


Average effect of allele substitution, cont.

Combining the two components


▶ average effect of substitution α is

α = p(a − d) + q(d + a) = a + d(q − p) = αi − αj

▶ this is the difference between the average effects of the two alleles
i = A and j = a

It turns out that


▶ αi = qα
▶ αj = −pα

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 32


Additive genetic value (A) or breeding value

Key concept
The αi values are the average effects of an allele.

Used to estimate the Breeding value or the additive genetic value


▶ key concept : parents only pass single alleles at each locus to their
offspring
▶ different offspring from the same parents will differ depending on the
allele that they receive from their parents
▶ parents pass the A part of their genotypic value G
▶ for 2 alleles, A = αi + αj
▶ for n alleles, A = nk=1 (α(k) (k)
P
i + αj )
▶ offspring A will be the average of the two parents A’s

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 33


Decomposing the genotypic values

Representing it graphically
Gij = GijA + δij = µG + αi + αj + δij

GAA ◦
Genotypic value

GAa ◦

Gaa ◦

0 1 2 Copies of allele A
aa Aa AA Genotypes
adapted from Bruce Walsh’s lecture

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 34


Decomposing the genotypic values

Representing it graphically
Gij = GijA + δij = µG + αi + αj + δij

GAA ◦
Genotypic value

GAa ◦

Gaa ◦

0 1 2 Copies of allele A
aa Aa AA Genotypes
adapted from Bruce Walsh’s lecture

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 34


Decomposing the genotypic values

Representing it graphically
Gij = GijA + δij = µG + αi + αj + δij


GAA ◦
Genotypic value

GAa ◦


Gaa ◦

0 1 2 Copies of allele A
aa Aa AA Genotypes
adapted from Bruce Walsh’s lecture

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 34


Decomposing the genotypic values

Representing it graphically
Gij = GijA + δij = µG + αi + αj + δij

• µG + 2αA
GAA ◦
Genotypic value

GAa ◦
• µG + αA + αa
µG + 2αa


Gaa ◦

0 1 2 Copies of allele A
aa Aa AA Genotypes
adapted from Bruce Walsh’s lecture

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 34


Decomposing the genotypic values

Representing it graphically
Gij = GijA + δij = µG + αi + αj + δij

• µG + 2αA
GAA ◦
Genotypic value

GAa ◦
• µG + αA + αa
µG + 2αa
α Slope = α = αa − αA

1

Gaa ◦

0 1 2 Copies of allele A
aa Aa AA Genotypes
adapted from Bruce Walsh’s lecture

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 34


Decomposing the genotypic values

Representing it graphically
Gij = GijA + δij = µG + αi + αj + δij

• µG + 2αA
GAA ◦ δAA
Genotypic value

GAa ◦
δAa
• µG + αA + αa
µG + 2αa
α Slope = α = αa − αA

1

Gaa ◦ δaa

0 1 2 Copies of allele A
aa Aa AA Genotypes
adapted from Bruce Walsh’s lecture

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 34


Population level data

Linear regression model


Gij = GijA + δij = µG + αi + αj + δij




GAA ◦


Genotypic value

GAa ◦




Gaa ◦

0 1 2 Copies of allele A
aa Aa AA Genotypes
adapted from Bruce Walsh’s lecture

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 35


Putting it into a linear regression

Adapting the representation


If we assume a biallelic locus
▶ let’s X1ij and X2ij be the number of alleles of type A and a, respectively,
(i.e. 0, 1 or 2), in genotype Gij
▶ as X1ij = 2 − X2ij , Gij becomes

Gij = β0 + β1 X1ij + δij

where
▶ β0 = µG + 2αj is the intercept
▶ β1 = αPj − αi is the average effect of allele substitution and the slope
▶ µG = (Gij × freqij )
▶ slope will change depending on the allelic effects, even with identical
genotypic values
▶ can be estimated in a linear regression of traits vs genotypes

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 36


Multi-locus traits

Polygenic traits, most common case


▶ alleles of these loci contribute to the genotypic mean (and thus the
phenotype)
▶ assume that the loci are unlinked (no epistasis, i.e. no interaction)
▶ population mean of the trait affected by L loci is simply
L
X L
X
µG = ai (pi − qi ) + 2 di pi qi
i=1 i=1

where, at loci i
▶ ai = genotypic value for the homozygotes with larger value
▶ di = genotypic value for the heterozygotes
▶ pi and qi are the two allele frequencies

Easily framed into a multiple linear regression too (see Zoltan’s course for
details)!

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 37


Exercices (from Falconer and Mackay, 1989)

Exercice 1
A locus is overdominant with the following genotypic values:
AA Aa aa
120 150 90
Find the allele frequency giving the maximum mean value. What is this
mean value? Tips: use a graph to plot the results and find the maximum

Exercice 2
What are the breeding values of the three genotypes above?

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 38


Exercices, cont.

Exercice 3
Use R to estimate the average effect of allele substitution and both allelic
effects in this example:
genotype<-c("aa","AA","Aa","AA","Aa","aa","AA",
"AA","AA","aa","aa","AA","Aa","AA","AA","aa",
"Aa","Aa","Aa","Aa","aa")
phenotype<-c(1.79,23.83,12.30,23.27,13.25,1.59,
23.14,23.70,25.09,1.82,2.40,22.94,12.51,24.13,
22.60,4.92,12.43,11.39,12.93,13.57,2.03)

Tips: transform the genotype and use lm() to get the intercept and slope.
Interpret the estimates as described in the course.

Assuming that the allele frequencies are fa = 0.43 and fA = 0.57, calculate
the allelic effects.

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 39


Further reading to do

Read Falconer and Mackay, 1989


▶ Chapter 7 – pages 111 to 124
▶ here’s the pdf
▶ ask questions through Teams

Nicolas Salamin – Introduction to Quantitative Genetics / Partie 1 – p. 40

You might also like