Genetics in Neuroscience – exam 2
GWAS and Interpretation of GWAS Results
Goals of disease genetics:
Predict if someone will get sick, how bad it will be and what treatment will work.
o Look into someone’s DNA and be able to say the chance to get a certain disease is
increased because of a certain genetic variant that you have.
o What treatment will work for this patient?
o What will the progression be of the disease?
Test hypotheses about relationships to other diseases and traits.
o Information about the disease is not always available in our DNA, there are also
environmental effects that can influence features of the disease.
o Might be some genetic variants that can predicts how fast a disease will progress / if
medicine will work.
o Is a disease genetically related to another disease?
o Predict comorbidity:
For example, if you have disease1 and there is some genetic predisposition
causing this disease, there could also be a higher risk of getting disease2.
If disease1 would be earlier, we could prevent disease2 from manifesting.
Understand the biology of the disease so we can design better treatments and diagnostics.
o Genetics stops at the p-value and cannot get any further.
o For this, we need to collaborate with neurobiologists.
Genetic variation: crucial for DNA to be informative for understanding individual differences. Without
genetic variation genes could not contribute to understand the risk of getting a disease. Use genes to
understand differences between individuals.
Twin / family studies: don’t directly measure the variation (/DNA) at all!
o They tell us something about the relationship between genetic individual and inform
us about the relative contribution of genes to the trait.
o Use known genetic relationships between individuals.
o Don’t sequence the DNA.
Association studies: measure all the genetic variation in the individuals being studied.
o If you have the genetic variation, you can test if this genetic variation contributes to
individual differences and contributes to getting a disease.
What creates genetic variation? increase genetic diversity
Mutation: introduces novel variants into the population.
o Spontaneous change in DNA in one individual offspring all succeeding
generations also have the mutation.
o In the population, with each generation, the mutation will become more frequent.
Recombination: re-shuffles the existing patterns of variation.
o Makes the combination of DNA different.
o Could make a difference inside a gene when recombination happens within a gene.
Consequences of mutation and recombination:
The genetic variants lying around the mutation / recombination site are transmitted to the
offspring together and are correlated! This correlation becomes smaller with every generation
because of recombination.
In the absence of recombination, this correlation would extend a great distance along
chromosomes.
Recombination breaks down this correlation over successive generations leaving a narrower
and narrower window of correlation between SNPs.
o Smaller blocks of correlated SNPs across each generation.
o D’ / R² are measures of correlation between SNPs.
1
, HAPMAP project: map the haplotype blocks in the human genome across different populations
(Europians / Asians, …). They genotyped parents-offspring (trios) and mapped the blocks of correlated
SNPs. These blocks are different across different populations due to different historical backgrounds.
Make a LD map
Insight into the organization of LD:
In every person a lot of DNA is the same, but some little blocks differ between people. At
these sites, there was probably recombination.
o Search for blocks with LD with limited haplotype diversity limited amount of
variance in this block.
In every LD block, there are a few variants possible.
o If there were a 1000 variants possible, then it wouldn’t be an LD block, because then
the SNPs in this block won’t be correlated.
Every color is one SNP (amino acid variant)
Orange = minor allele (less frequent)
Blue = major allele (more frequent)
The HAPMAP project found that in LD blocks, there are only
a few variants possible. So, if you know SNPs in this
block, you could predict with variant the person has!
Example LD block: 4 different variants.
If the R²=1 for two SNPs (first two) then you can always predict which
SNP variant this person has on SNP2 based on the variant of SNP1.
For this block, you only need to genotype SNP1, SNP3 and SNP4 to
know the whole genotype if this LD block, because they are perfectly
correlated (predictive) with the other 3 SNPs.
Actually, in this case you only have to genotype SNP1 and SNP3 in order to
know the whole genotype.
You always genotype on the +strand of the DNA. In this example, the variant of SNP1 on the
+strand can be either A or T.
Due to the HAPMAP project we know which SNPs are informative for the whole genotype of
the LD block, so which SNP we have to genotype.
This saves a lot of money!
Use LD to impute missing genotypes:
Apart from not needing to genotype all variants, correlation between SNPs also allows us to fill
in un-genotyped SNPs.
This allows to meta-analysis data from different studies that have used different genotyping
arrays.
We use LD for:
Designing GWAS arrays, saves a lot of money.
Impute the SNPs that we did not genotype (predict).
2
GWAS and Interpretation of GWAS Results
Goals of disease genetics:
Predict if someone will get sick, how bad it will be and what treatment will work.
o Look into someone’s DNA and be able to say the chance to get a certain disease is
increased because of a certain genetic variant that you have.
o What treatment will work for this patient?
o What will the progression be of the disease?
Test hypotheses about relationships to other diseases and traits.
o Information about the disease is not always available in our DNA, there are also
environmental effects that can influence features of the disease.
o Might be some genetic variants that can predicts how fast a disease will progress / if
medicine will work.
o Is a disease genetically related to another disease?
o Predict comorbidity:
For example, if you have disease1 and there is some genetic predisposition
causing this disease, there could also be a higher risk of getting disease2.
If disease1 would be earlier, we could prevent disease2 from manifesting.
Understand the biology of the disease so we can design better treatments and diagnostics.
o Genetics stops at the p-value and cannot get any further.
o For this, we need to collaborate with neurobiologists.
Genetic variation: crucial for DNA to be informative for understanding individual differences. Without
genetic variation genes could not contribute to understand the risk of getting a disease. Use genes to
understand differences between individuals.
Twin / family studies: don’t directly measure the variation (/DNA) at all!
o They tell us something about the relationship between genetic individual and inform
us about the relative contribution of genes to the trait.
o Use known genetic relationships between individuals.
o Don’t sequence the DNA.
Association studies: measure all the genetic variation in the individuals being studied.
o If you have the genetic variation, you can test if this genetic variation contributes to
individual differences and contributes to getting a disease.
What creates genetic variation? increase genetic diversity
Mutation: introduces novel variants into the population.
o Spontaneous change in DNA in one individual offspring all succeeding
generations also have the mutation.
o In the population, with each generation, the mutation will become more frequent.
Recombination: re-shuffles the existing patterns of variation.
o Makes the combination of DNA different.
o Could make a difference inside a gene when recombination happens within a gene.
Consequences of mutation and recombination:
The genetic variants lying around the mutation / recombination site are transmitted to the
offspring together and are correlated! This correlation becomes smaller with every generation
because of recombination.
In the absence of recombination, this correlation would extend a great distance along
chromosomes.
Recombination breaks down this correlation over successive generations leaving a narrower
and narrower window of correlation between SNPs.
o Smaller blocks of correlated SNPs across each generation.
o D’ / R² are measures of correlation between SNPs.
1
, HAPMAP project: map the haplotype blocks in the human genome across different populations
(Europians / Asians, …). They genotyped parents-offspring (trios) and mapped the blocks of correlated
SNPs. These blocks are different across different populations due to different historical backgrounds.
Make a LD map
Insight into the organization of LD:
In every person a lot of DNA is the same, but some little blocks differ between people. At
these sites, there was probably recombination.
o Search for blocks with LD with limited haplotype diversity limited amount of
variance in this block.
In every LD block, there are a few variants possible.
o If there were a 1000 variants possible, then it wouldn’t be an LD block, because then
the SNPs in this block won’t be correlated.
Every color is one SNP (amino acid variant)
Orange = minor allele (less frequent)
Blue = major allele (more frequent)
The HAPMAP project found that in LD blocks, there are only
a few variants possible. So, if you know SNPs in this
block, you could predict with variant the person has!
Example LD block: 4 different variants.
If the R²=1 for two SNPs (first two) then you can always predict which
SNP variant this person has on SNP2 based on the variant of SNP1.
For this block, you only need to genotype SNP1, SNP3 and SNP4 to
know the whole genotype if this LD block, because they are perfectly
correlated (predictive) with the other 3 SNPs.
Actually, in this case you only have to genotype SNP1 and SNP3 in order to
know the whole genotype.
You always genotype on the +strand of the DNA. In this example, the variant of SNP1 on the
+strand can be either A or T.
Due to the HAPMAP project we know which SNPs are informative for the whole genotype of
the LD block, so which SNP we have to genotype.
This saves a lot of money!
Use LD to impute missing genotypes:
Apart from not needing to genotype all variants, correlation between SNPs also allows us to fill
in un-genotyped SNPs.
This allows to meta-analysis data from different studies that have used different genotyping
arrays.
We use LD for:
Designing GWAS arrays, saves a lot of money.
Impute the SNPs that we did not genotype (predict).
2
