genome evolution. amos tanay 2010 genome evolution lecture 10: quantitative traits
TRANSCRIPT
Genome Evolution. Amos Tanay 2010
Genome evolution
Lecture 10: Quantitative traits
Genome Evolution. Amos Tanay 2010
Every meaningful evolutionary traits is ultimately quantitative
F. Galton
Continuous traits: Weight, height, milk yeild, growth rate
Categorical traits: Number of offspring, petals, ears on a stalk of corn
Threshold traits: disease (the underlying liability toward the trait)
Ultimately, fitness is a quantitative trait, so what is special about it?
Historically, research on genetics and directed selection were distinct from evolutionary theory
Currently, a quantitative approach to molecular evolution and population genetics is a major frontier in evolutionary research
Genome Evolution. Amos Tanay 2010
The basic observation: heritability
22 )()()( xExExVar
)()()(),( yExExyEyxCov
2)]([ bxayESS
A linear fit would try to minimize the mean square deviation:
0)]()()([2))((2
0))()((2)(2
2
xbExaExyEbxayxEb
SS
xbEayEbxayEa
SS
)()(
),(
)(
),(
0])()([)]()()([ 22
yVarxVar
yxCovr
xVar
yxCovb
xExEbyExExyE
2
2
1hb
Heritability is defined:
(dividing because only one parent is considered)
This is the “narrow sense” heritability
Genome Evolution. Amos Tanay 2010
Artificial selection
Over 100 years of an ongoing selection experiments
12
12
48
%O
il
From 4.6% to 20.4% oil
What kinds of evolutionary dynamics allow for such rapid increase in the trait?
Genome Evolution. Amos Tanay 2010
Artificial selection
Selection can work by exploiting existing polymorphic sites
SNP data suggest that at least 50 genes were involved in the corn selection
or by fixating new mutations
Theory suggest that fixation of all strong effects should occur rapidly – 20 generations. Later one should see fixation of alleles with smaller effect or new mutations
)2ln()/2( Nst Remainder- Theorem (Kimura):
One strong candidate for introducing mutations are repetitive elements.
The corn population is of tiny size (60)
Selection is enhanced due to the threshold effect
Genome Evolution. Amos Tanay 2010
Limits to artificial selection
After some (variable number of) generations, artificial selection stop increasing the trait
One reason for that can be the exhaustion of polymorphismThis is frequently not the case, since reversing the selection is frequently shown to have an effect – meaning polymorphisms is present
Another reason for converging trait values is selection on other traits (fertility!)
Using many allele affecting the trait, artificial selection can reach trait values that are practically never observed in the original population
Not all traits can be artificially selected: in 1960, Maynard-Smith and Sondhi showed they could not select for asymmetric body plan in flies by choosing flies with excess of dorsal bristles on the left side
This suggest that some traits are strongly stabilized
Artificial selection can proceed non-linearly: starting and stoppingA main possible reason for that is that recombination of strongly linked alleles takes time
J. Maynard-Smith
Genome Evolution. Amos Tanay 2010
Truncation selection
M
MS
S = MS - M
R = M’ - M
M’
Response:
Differential:
The selection differential is generally larger than the selection response
This is because some of the selected offsprings are of high trait value due to non-genetic effects
Another reason is that the genotype of the selected offspring is modified by segregation and recombination
ShR 2
We redefine (realized) heritability as the ratio between selection differential and selection response
Genome Evolution. Amos Tanay 2010
Back to genetics: two loci
4
3
2 1 0
2
23
1
AA
Aa
aa
BB Bb bb
MS = 0.8Selecting class 0 and 1
1/161/16
1/16 1/16
2/16
2/16
2/16
2/16 4/16p(A)=p(B) = 0.2After selection:
M’ = 0.8Yielding:
Generally: M=2(p(A)+p(B))
Assume additive selected trait
h2 = 1and
4
4
2 2 0
4
24
2
AA
Aa
aa
BB Bb bb
MS = 12/7Selecting class 2,0
1/161/16
1/16 1/16
2/16
2/16
2/16
2/16 4/16
p(A)=p(B) =2/7After selection:
M’ =96/49Yielding:
Assume dominant selected trait
h2 =17/21=0.81and
Genome Evolution. Amos Tanay 2010
Continuous traits: regression of alleles and phenotypes
M
MS
BZ
T
(MS-M)/
AA
AA’
A’A’
2''AAAA mm
m
2''AAAA mm
a
mmd AA '
We now assume each genotype have a distribution of trait values
The variability may be a consequence of environmental factors or other loci
m – meana – additivelyd = dominance
ammAA
amm AA ''dmmAA '
pqdaqpm
amqdmpqampM
2)(
)()(2)( 22
dpqpqpqa
ppqdaqpmpqdappm
)(22
)2](2)([222 2
Cov(pheno, number of A alleles)=
Var(number of A alleles)= pqppqp 2)2(24 22
dpqapq
dqppqpqab )(
2
)(22
Genome Evolution. Amos Tanay 2010
Continuous traits: truncation selection
M
MS
BZ
T
(MS-M)/
AA
AA’
A’A’
2''AAAA mm
m
2''AAAA mm
a
mmd AA '
ammAA
amm AA ''
dmmAA '
aT dT
aT
Selecting on a threshold over the mean of the population (T)
Thresh relative to AA normal distrib
Thresh relative to AA’ normal distrib
Thresh Relative to A’A’ normal distrib
The “fitness” equals the ratio between the areas beyond the threshold
Assuming small differences, the areas are nearly rectangular:
)())()((1211 daZaTdTZww
)())()((2212 daZdTaTZww
Genome Evolution. Amos Tanay 2010
Allele frequency change
M
MS
BZ
T
(MS-M)/
AA
AA’
A’A’
2''AAAA mm
m
2''AAAA mm
a
mmd AA '
ammAA
amm AA ''
dmmAA '
It can be generally shown (compute!) that:
)())()((1211 daZaTdTZww
)())()((2212 daZdTaTZww
wwwqwwppqp /)]()([ 22121211
BdaqZdapZpqp /)]()([
Average fitness is the area B:
])([)/( dpqapqBZp
SelectionIntensity
Allelefrequency
Phenotype to genotyperegression
Genome Evolution. Amos Tanay 2010
Mean Phenotype
M
MS
BZ
T
(MS-M)/
AA
AA’
A’A’
2''AAAA mm
m
2''AAAA mm
a
mmd AA '
ammAA
amm AA ''
dmmAA 'pdpqaM
ppqaamqdmpqamp
ampqdmpqppamppM
])([2
)]([2)()(2)(
)()())()((2)()('22
22
])([)/( dpqapqBZp
SelectionIntensity
Allelefrequency
Phenotype to genotyperegression
22
22
2
/])([2
/])([2)(
])([2)/(
])([2'
dpqapqSR
dpqapqMM
dpqapqBZ
pdpqaMM
S
222 /])([2 dpqapqh
Interpretation: Proportion of phenotypic variance that can be explained by genotype change
Genome Evolution. Amos Tanay 2010
Heritability
Whither height
%protein in milk
Feed efficiency
Milk Yield
Calving interval
Cattle
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Egg weight
Body weight
Albumen cont.
Sexual maturity
Eggs/hen
Poultry Swine
Litter size
Daily gain
Body length
Back fat
Feed efficiency
Yield
Ear number
Ear weight
Plant height
Maize
Nar
row
sen
se h
erita
bilit
y
Human: Stature – 0.85 Weight – 0.62 Handedness – 0.31 Fertility – 0.1-0.2
Genome Evolution. Amos Tanay 2010
“..What have we learned about the nature of quantitative trait variation for height from these studies? At a first glance it looks quite simple: variation is explained by many variants of small effects, with no evidence for interactions between alleles, either within loci (dominance) or between loci (epistasis), and there are no strong differences in effects between males and females. These observations are consistent with patterns of familial resemblance for height. However, given the design and analysis used, there was little statistical power to find evidence for departures from this simple model. Not surprisingly, given the small effect sizes found, there was no significant overlap between the location of the associated variants and previously reported loci from linkage studies. It remains a challenge to reconcile the findings of GWAS and linkage studies, because the former suggest individual variants with small effects, whereas the latter suggest genomic regions with large effects within pedigrees. “
Genome wide association studies for Stature
GWAS use large cohort (63,000 in this example) of unrelated individuals..Simulations of detection power:
Standard approach extend what we have shown above to small pedigrees and look for linkage of QTL
Genome Evolution. Amos Tanay 2010
Genetic analysis of genome-wide variation in human gene expression (Morely et al. 2004)
14 CEPH families (of ~8 members each)
3554 variable expression genes (in lymphoblastoid cells)
2756 SNPs (just a few!)
Alternatively: 94 unrelated CEPH grandfathers
Testing linkage of expression and SNPs in the large family trees yield linkage for ~1000 phenotypes
The test on families use the genealogical structure (SIBPAL - http://darwin.cwru.edu/)
Alternative test on unrelated individuals use simple correlation of the 0,1,2 individual
Difficulties: multiple testing vs low resolution
Reporting on loci that are linked with many QTLs
Genome Evolution. Amos Tanay 2010
Variability in B-cells response to irradiation
Mapped eQTL15 CEPH families (of ~8 members each)
Low resolution ~3000 SNPS, and high resolution HapMap SNPS,
3280 responding genes – different time points during irradiation
Follow up molecular biology experiments
Genetic analysis of radiation-induced changes in human gene expression (Smirnov 2009)
Genome Evolution. Amos Tanay 2010
Genetic Dissection of Transcriptional Regulation in Budding Yeast (Brem et al 2002)
Crossing two budding yeast strains
Fully genotyping, testing expression (later in different conditions)
Hundred of variably expressed genes
Using the compact yeast genome help deciding linkage
Using the well-characterized biology of yeast helps explain linkage
Genome Evolution. Amos Tanay 2010
Building association to groups of genes instead of single genes (Litvin et al 2009)
©2009 by National Academy of Sciences
Identifying regulatory mechanisms using individual variation reveals key role for chromatin modification (Lee et al 2008)
Genome Evolution. Amos Tanay 2010
Schadt EE et al. 2005 (and many publications following it)
R – expressionL – locus genoetype
C - phenotype
Looking for gene expression traits that explain QTLs – stands between genetic loci and some disease trait of interest
Applied to obesity linkage (in mice)
Further development use more data (not just expression), or gene subnetworks
Ultimate goal is to build a model explaining phenotype by genotype through molecular phenotypes
Possible modes of causality or interaction
Positive correlation suggests linked eQTLs
Correlation between genetic distance and correlation suggests LD effect
Genome Evolution. Amos Tanay 2010
(a) The horizontal axis shows the frequency from the NMR spectrum expressed as chemical shift from right to left ( , ppm). The vertical axis indicates genetic locations (cM) on chromosomes 1 to X. The lod scores between each genotype and each metabolite are color coded. Significant linkages between genomic locations and regions of the plasma NMR profile are present in the aliphatic region (0.5 to 4.5 ppm) and the aromatic region (>5.5 ppm). Resonances corresponding to the anesthetics and their degradation products were withdrawn as described in Methods. (b,c) Genome-wide linkage mapping across the full metabonomic spectrum for marker D14Wox10 (b) and linkage data across the genome for the metabolite 7.86 (c).
Direct quantitative trait locus mapping of mammalian metabolic phenotypes in diabetic and normoglycemic rat models (Dunas et al. 2007)
Crossing two rat strains: diabitic and normal
2000 microsatellite and SNP markers
Using NMR to perform metabolic profiling – looking for linkage explaining metabolic abnormalities
Genome Evolution. Amos Tanay 2010
JK Pickrell et al. Nature 000, 1-5 (2010) doi:10.1038/nature08872
Understanding mechanisms underlying human gene expression variation with RNA sequencing
Genome Evolution. Amos Tanay 2010
Loci affecting isoform expression.
“..we identified more than a thousand genes at which genetic variation influences overall expression levels or splicing..”
Genome Evolution. Amos Tanay 2010
Phenotype variation: back to evolutionary theory
Phenotypes in natural environment can be modeled as a combination of genotype and environmental effects:
EGP
More carefully, the genotype effect on phenotype may is a function of the environment, and the additive form may be wrong
For example, gene expression of stress related genes depends on the genotype differently for different stresses
Understanding QTL evolution
Mapping phenotypes to QTL