GCD exam 2

haplotypes

are combinations of alleles along a physical chromosome Only some of the possible haplotypes actually exist in a population Haplotypes reflect "linkage disequilibrium" We can use "tag" SNPs to infer which haplotype a person carries

Paradigm for communication between environment and genome

ligand/ morgphogen -> receptor -> inhibitors/ signal transducers -> trans-acting factors -> target genes the genes that morphogens turn on and off are typically transcription factor genes, which then regulate other genes

Finding genes that cause disease

linkage analysis or association analysis

Linkage mapping strengths and weaknesses vs association analysis

linkage: good for related individuals with rare dominant traits • Less commonly done • Useful for rare variants • Larger and larger pedigrees association: can be used for related and nonrelated individuals for common traits

Pax6

loss of function leads to Aniridia (human) and small eye (mouse) controls eye development. heterozygotes have very small eyes, homozygotes have no eyes

Possible effects of decreasing effective level of PAX3

100% Normal development >50% Dystopia canthorum 50% Patchy failure of melanocytes <50% Limb muscle defects <10% Neural tube defects The frontal bone is the most sensitive to PAX3 decrease. Differentiation and survival of melanocytes is less sensitive to decreases in PAX3. Development of limb buds is relatively insensitive and normally affected only in homozygotes for PAX3 loss of function variants.

Gene-Environment Interaction in Development

It has been demonstrated that administration of ethanol to chick embryos resulted in a dramatic loss of Shh, as well as a loss of transcripts involved in Shh signaling pathways. In contrast, other signaling molecules examined did not demonstrate such dramatic changes. Furthermore, it was demonstrated that both ethanol induced cranial neural crest cell death and the associated craniofacial growth defect can be rescued by application of Shh. These data suggested that craniofacial abnormalities resulting from fetal alcohol exposure are caused at least partially by loss of Shh activity and subsequent neural crest cell death.

Relatedness among family members is confounded with shared environment

• Family members share alleles. • But they also have more similar environments. • We need to separate the genetic influences from the shared environment. twin studies • Monozygotic (MZ) twins arise from cleavage of a zygote => their genomes are virtually identical • Dizygotic twins (DZ) arise when two eggs are fertilized at the same time => related just like non-twin siblings • MZ and DZ twins share different amounts of alleles, but share the same environment • Allows separation of environmental & genetic effects

parent offspring regression

• For quantitative traits (e.g. blood pressure, height), we can measure "familial aggregation" via correlations among relatives • The more correlated trait values are between relatives, the more evidence we have for shared (genetic) factors

genetic markers

• Genetic "markers" are DNA variants in a population or pedigree • We can assay these markers to determine the alleles carried by each person • We can then check which markers are linked to a given disease • Typically, we do NOT think that the markers are causing disease! They are just close to the causal genes. • There are two major types of markers: variable length polymorphisms (# short tandem repeats), and SNPs

Are there genetic human races?

• Genetically, humans do not fall into clear, discrete categories that align with historic "race" categories • Human genetic variation does reflect our ancestry, which differs by the geographic region in which our ancestors lived • Genetic ancestries form geographic gradients that reflect our population history in 1) subtle frequency shifts of common alleles and 2) the presence of locally common rare variants

some important histone marks

• H3K4me3: an "active" mark. Usually in active promoters. Most other me3 marks are repressive! • H3K4me1: intergenic, in enhancers (both active and not) • H3K27me3: in repressed regions of the genome • H3K27ac: strongly active, in promoters or enhancers • H3K36me3: represses transcription behind an RNA polymerase; indicative of active transcription • H3K9me3: represses repeat regions

Twin studies and "concordance"

• High concordance between MZ twins suggests a genetic basis. • MZ concordance less than 100% indicates environmental factors. • If MZ and DZ concordance are similar (even if both are high), the trait is less likely to be genetic. • Higher concordance in MZ than in DZ twins indicates a genetic basis. The larger this difference, the more "genetics" there is. Complex diseases are due to both genetic and environmental factors. The genetic component is responsible for genetic susceptibility. The disease occurs when a genetically susceptible individual is exposed to the appropriate environmental factors.

Variant calling

• Illumina reads have 1 error every 1,000 bp (on average) • Count the number of times you see the reference allele and alternative alleles • Calculate the likelihood of observing these reads under two models: • 1. "there is a variant here" • 2. "there is no variant here" [i.e., the apparent allele is actually a sequencing error] • Assign "genotype quality scores"

Hardy-Weinberg equilibrium

• Relates allele frequencies to genotype frequencies: • Allele frequencies: p + q = 1 • Genotype frequencies: p2 + 2pq + q2 = 1 • HWE holds under certain assumptions: • Large population (random events, genetic drift) • Random mating (cultural influences on the genome) • Constant allele frequencies: • no new mutations (new mutations create diversity) • no selection (adaptations) • no migration (population mixture) • Things get interesting when these assumptions are NOT met!

DHODH is the causal gene for Miller syndrome

• Sanger sequencing in four additional cases • All cases are compound heterozygotes for "deleterious" missense variants • Parents each heterozygous for one variant • Recessive disorder; not caused by new mutations • Causal mutations rare in the population • Biology of disease still unclear, but now potential link to other disease with known metabolic basis

Genetic drift is influenced by other factors

• Selection occurs when a variant influences fitness. Selection can be positive ("adaptive") or negative ("purifying", "stabilizing") • Migration occurs when populations exchange individuals (and their genetic variants) • Linkage ties the fate of neighboring variant to each other • Population history (expansions, bottle necks) can drastically influence the genetic makeup of a population

Mutation rates for other types of mutations

• Some types of mutations happen more often than those at single nucleotides: - Mitochondrial genomes: 5-10 fold higher than nuclear - Microsatellites: up to 7 x 10-3 - CpG dinucleotides: 10x higher than average - Certain labile sites in the genome - Large structural mutations? • These differences require gene-specific models of mutation rates when calculating the "expected" number of variants in a given gene in the population

Exceptions to random mating

• Stratification describes a population with multiple subgroups that have remained relatively separate due to historical, cultural, or religious reasons. • The U.S. remains stratified by ethnicity; India is stratified by caste • Assortative mating is the choice of mate due to a particular trait. People tend to choose mates who resemble themselves (e.g. in native language, appearance, talents, interests, ...) • Consanguinity in genetic isolates (e.g. on small islands) or due to religious isolation. Consanguinity can result in enrichments of otherwise very rare alleles (e.g. Tay-Sachs disease has 1/30 carriers in Ashkenazi Jews in North America vs. 1/300 in non-Ashkenazi). • Each of these phenomena exists in human populations, and distorts genotype frequencies relative to HWE, for example resulting in more homozygotes than expected. • They do not change allele frequencies.

Heritability

"Heritability" measures the genetic vs. environmental contribution to a trait phenotype = genes + environment P = G + E • ...is the proportion of variance ("V") in a trait that is due to genetic variation. • H2 = Vgenetic / Vtotal • Typical H2 values range from 0.2 for some cognitive traits to 0.8 for some morphological traits (e.g. height). • Heritability is specific to a given population • Heritability can change over time • Even without any change in genetic variation, heritability can change if the environment changes • A 100% heritable trait can be changed by the environment • Strong effect mutations (e.g. in BRCA1) can produce small heritability • High heritability does not imply genetic determinism

Analysis of next-gen sequencing data

Data cleaning - Exclude poor base calls - Reduce bias due to PCR amplification (some starter molecules get copied many times) Alignment to reference sequence - repetitive regions are difficult to align Genotype calling - "Is there a variant here?" - "What is this person's genotype?" - Requires multiple-fold "coverage" for confident genotype calls

Genome Wide Association Studies (GWAS)

GWAS (genome wide association studies) examine genetic variation across the entire genome, covering nearly all common variants. GWAS permits rapid scanning of markers across the genome to find which variants are associated with a particular disease. Generally include two populations: one with ("cases") and one without ("controls") the condition of interest. "Conditions of interest" may include disease, biomarkers, response to drug therapy, etc.

Histone mark naming rules

H3K4me3 H=histone 3 = what histone which amino acid (K = lysine) 4 which lysine on the histone methylation (me) or acetylation (ac) how many methyl groups (only for methylation)

gene expression in different tissues

In most tissues, a few hundred genes account for 50% of total expression • Often, most expression is mitochondrial (e.g. kidney) • In blood, three hemoglobin genes make up 60% of total transcription • Brain and testis express the most genes

Beyond mRNA: molecular traits

Just like for mRNA, we can measure genetic influences on other "molecular traits", for example: • Chromatin (DNA methylation, DNAse hypersensitive sites, CHIPSeq) • Gene expression dynamics (mRNA transcription & degradation) • Splicing (appears to be a major contributor to common disease) • mRNA translation • RNA modifications, transcription start sites, polyA-sites

causes of dysmorphology

Known genetic transmission 20% Chromosomal transmission 3-5% Environmental causes: - Radiation <1% - Infections 2-3% - Maternal factors 1-2% - Drugs/environment 4-6% Unknown 65-70%

genetics of BCNS

Linkage studies indicated that the gene for Basal Cell Nevus Syndrome (BCNS) was on chromosome 9q22. This is the same location for the homologue of patched in Drosophila development (PTCH1 in humans). Multiple mutations, usually truncating mutations, were found in the PTCH1 gene in humans associated with BCNS

Patched gene

Patched is a strong candidate as a tumor suppressor gene. The classic Knudson (two-hit) hypothesis can then explain the neoplastic process. The inactivation of both copies of a tumor suppressor gene leads to tumor development. In the case of patched where one mutation may be inherited, evidence currently points to UV-induced DNA damage as the main culprit in patched inactivation. Jaw cysts are also thought to be due to loss of heterozygosity (LOH)

Gorlin Syndrome molecular mechanism

Patched, the receptor for Sonic Hedgehog, normally suppresses Smoothened. If Patched is mutated, suppression no longer occurs and message goes to nucleus and activates gene expression via the Ci / Gli transcription factors

Qualitative vs. Quantitative traits

Qualitative traits Usually a single gene trait. Either you have the disease or you do not. ~ 20% of traits, exhibited in early childhood Quantitative traits Phenotype is usually a continuum, such as the risk for a common disease. Multiple genes and environmental factors play roles. Individuals can have "disease genes" and not have the disease. ~80% of traits, delayed onset

Three important numbers in statistical testing

Significance: Is an association between SNP and disease real? Effect size: How strong is the effect of the variant? Power: How likely are we to find a real association?

Synpolydactyly (SPD)

Synpolydactyly (SPD) is a dominantly inherited congenital limb malformation. Typical cases have 3/4 finger and 4/5 toe syndactyly, with a duplicated digit in the syndactylous web, but incomplete penetrance and variable expressivity are common. Synpolydactyly (SPD) caused by polyalanine tract expansions in HOXD13

dysmorphology

The study of abnormal development where physical features are unusual and can provide a clue to pathogenesis

developmental genetics

The study of normal and abnormal development - how cellular diversity and specificity is generated through gene expression and developmental cell signaling

signal cascade during early development

maternal effect genes -> gap genes -> pair rule genes -> segment polarity genes signals from one leads to signals from the next

Drosophila bicoid

the first morphogen to be discovered binds to DNA and RNA In Drosophila: Bicoid influences the transcription of genes that create the anterior sections of the organism (the head). Wherever the bicoid gene is expressed, the head will develop.

summary of regulatory genetic variation

• Most GWAS hits are non-coding & act by altering gene expression • Almost every gene has one or more eQTLs • Almost every common human SNP is an eQTL • We now have a few (but only a few) GWAS regions for which we have linked specific noncoding variants to a complex disease • Causal genes are oten far from the variant, and not easy to guess

Waardenburg Syndrome

•Autosomal dominant neurocutaneous syndrome •Congenital sensorineural hearing loss •Patchy hypopigmentation (white forelock) •Distinctive facial features Dystopia canthorum (eyes appear wider apart) Broad and high nasal root (top part of nose)

height

Height varies a lot Height is quantitatively distributed • Height has a heritability of ~80% • It is one of the most highly studied human traits • (because it can be easily measured when conducting GWAS for other traits) • For some loci, an association with height can be explained by gene function • For many other loci, the connection between gene and trait is not straightforward lots of variants (800 identified)

Illumina sequencing

Illumina sequencing is based on "reversible dye-terminators". Four types of fluorescently labeled bases are added to the reaction chamber. The correct base is chemically bound to the new DNA strand using DNA polymerase, while nonincorporated nucleotides are washed away. The DNA is extended one nucleotide at a time. A camera takes an image of the fluorescence-labeled growing DNA strand. Then, the dye along with the terminal 3' blocker is chemically removed from the DNA, allowing the next cycle. Basically, a microscope & camera in a PCR machine, looking at a millions of fluorescent DNA molecules as they grow. 6 trillion bases in about 40 hours 2,000 human genomes ~67 human genomes to 30X coverage Many other potential uses • Results in hundreds of millions of reads • Reads reads are "short", i.e. 50 -250 bp • Reads can be single-end or paired-end • We can detect variants in individual genomes

Where is DNA methylation in the human genome?

Primarily at CpG dinucleotides (over 80% methylated in somatic cells) • Deaminated 5mC is a "T" => mismatch repair will often turn 5mC into T • CpG dinucleotides are underrepresented in vertebrate genomes • Those that remain reside primarily in un-methylated "CpG islands" • CpG islands are enriched in promoters of most "housekeeping" genes • Reduced methylation can point to presence of a promoter or enhancer

effects of histone marks

• Can be direct: e.g. acetylation loosens chromatin structure because it removes a positive charge at lysine, reducing the affinity to DNA • Often indirect, by attracting other factors. E.g. some me3 can attract HP1, a protein involved in establishing and spreading heterochromatin • An enormous number of possible combinations of marks • Some combinations are seen more than others

Who has DNA methylation?

• Human? Yes • Mice? Yes • Yeast? No • C. elegans? Little • Drosophila? Little • Plants? some • Bacteria? some DNA methylation can affect transcription factor binding

Genetic basis of even more complex traits

• Many cognitive traits and neuropsychiatric diseases have extremely polygenic architectures • Few variants have been identified so far, and they have extremely small effects • This might be because larger effects on brain function would be selected against by evolution, leaving only very small effects left in the population.

Genetics of complex trait take aways

• Many variants, typically with small effects • Larger GWAS & sequencing studies identify more and more variants • Identifying the causal genes and variants in a GWAS region is challenging, and one of the major current areas of human genetic research • The variants can provide insight into trait biology and suggest pathways for targeting by drugs

Two ways to quantify effect size of associations

• Odds ratio (OR) • used in case-control designs • Relative risk (RR) • used in cohorts and general population samples • OR and RR are closely related, but used in different contexts

genetic mapping using linkage

General Concept: Identify marker alleles that track the disease. Look for "co-segregation" of the disease phenotype with a marker allele. We want to find genetic markers that are located close to the disease gene.

linkage disequilibrium

Linkage disequilibrium (LD) is the non-random association of two or more alleles. (the opposite, linkage "equilibrium" would occur if all variants were unlinked and fully independent of each other. GWAS would then require many more markers) r2 ranges between 0 and 1: • 1 when the two markers are in complete LD • 0 when they are in complete equilibrium (random segregation) When r2 = 1 the two markers provide exactly the same information r2 is a correlation coefficient Because of their relatively high allele frequency, it is usually assumed that a marker variant on a genotyping array is not itself functional (pathogenic). When a marker allele is associated with a disease, we assume that this marks a DNA segment that also contains a different disease producing mutation that sometimes is in the same segment. The "risk allele" of the associated SNP marker will not always be associated with the disease producing mutation. "Linkage Disequilibrium" between a genetic marker and a causal variant (functional mutation) Linkage Disequilibrium connects "neutral" markers with causal variants in functional elements

Methylation

Methylation at CpG dinucleotides (C to 5mC) DNA Methylation can be inherited across cell division DNMT3A and DNMT3B DNA methyltransferase 3A and 3B de novo methylation DNMT1 (DNA methyltransferase 1) performs "maintenance methylation" on hemi-methylated DNA The pattern of DNA methylation is heritable by somatic cells and maintained after DNA replication by DNA "maintenance methyl transferase", which has a 100-fold greater affinity for hemi-methylated DNA (i.e., parent strand methylated, daughter strand un-methylated) than for un-methylated DNA. removing methylation • Simple diffusion across mitosis without maintenance methylation? • TET enzymes actively remove 5mC

Odds Ratio (OR)

OR = odds of disease with allele / odds of disease without allele The Odds Ratio (OR) can be used to determine how likely an individual with the risk allele will have the specific phenotype. • "What are my odds of getting the disease if I have this allele, relative to if I do not?" • The OR gives an indication of the effect size that a given allele has on the phenotype. • An OR of 2.0 for the risk allele means that there is a 2 fold higher likelihood that an individual who inherits the risk allele will get the disease when compared to individuals without the risk allele. • A confidence interval provides a likely range of the OR. The more narrow the range, the better the estimation of the effect size. • Individuals who inherit the risk allele do not always have the phenotype • Individuals can get the disease without the risk allele odd = case with allele / control with allele (a/b) or case without allele/ control without allele (c/d)

relative risk (RR)

RR = [a/(a+b)]/[c/(c+d)] a= cases with allele b = control with allele c = cases without allele d = control without allele In a comparison between an experimental group and a control group: A relative risk of 1 means there is no difference in risk between the two groups. An RR of < 1 means the event is less likely to occur in the experimental group than in the control group. An RR of 0.5 means the risk is ½ that of the average An RR of > 1 means the event is more likely to occur in the experimental group than in the control group. An RR of 2 means the risk is twice that of the average

Relative risk λr for discrete ("yes or no") traits

RR = prevalence of the disease in a relative of an affected person / population prevalence of disease A relative risk (RR) of 1 means there is no difference in risk between the two groups. An RR of > 1 means the event is more likely to occur in the related group than in the control group. The assumption is that increased λr is caused by the alleles we share with relatives. Large λr indicates that genetics plays a large role in these diseases.

Exome sequencing

The "exome" is the set of all human exons (but not the introns, and no intergenic sequence) The exome is ~30 Mb in size (~1% of genome) Much less than the whole genome! Variation in protein coding exons is easier to interpret than in non-coding regions

bonferri correction

The Bonferroni correction is a statistical method used to counteract the problem of multiple comparisons. In GWAS, we perform a large number of tests (i.e. 1 x 106 SNPs). At p-value ≤ 0.05, there will be on average 50,000 false positives! To control for this, we use the Bonferroni correction to set a more stringent p-value: pBonferroni ≤ 0.05 / the number of tests For 1 million tests, this gives a cut-off for significance of 0.00000005 (5 x 10-8) To reach such levels of significance, large sample sizes are required, or weak associations will be missed. For many common diseases, the effects of each risk allele is very small, requiring very large study populations.

Absolute Risk vs. Relative Risk

The population risk for asthma is 7% Allele gives you a RR of 2.0 Absolute risk for individual is ... ? 14% The population risk for multiple sclerosis is 0.3% Allele gives you a RR of 10 Absolute risk for individual is 3% (97% disease free) A large relative risk can still be small in absolute terms!

recombination fraction

The recombination fraction (θ, "theta") measures the degree of linkage. It ranges from 0 to 0.5. θ is the fraction of meioses in which two markers are separated by recombination in a pedigree. Two loci are said to be linked when θ < 0.5. When θ = 0.5, then loci (or marker and phenotype) are segregating independently and there is no linkage.

Using exome sequencing for disease gene discovery

Two strategies: 1. Follow-up on linkage mapping - sequence everything in the linked region 2. For traits caused by new mutations (e.g. severe autosomal dominant traits that cannot be passed on): do not "map" at all. Instead, just sequence the patient and their relatives, and directly look for new mutations that cosegregate with the trait

Direct estimation of the human mutation rate

Whole-genome sequencing of parent-offspring trios found that a newborn inherits ~70 spontaneous new single nucleotide mutations on average. The estimated mutation rate for humans is 1.2 x 10-8 mutations per generation per nucleotide. There are so many humans on the planet (more people than bases in the genome) that each single base mutation that is compatible with life probably exists in at least one person. Older fathers produce more mutations (an additional 2.01 mutations/year)

epigenetics:

first epigenetics: • In the 1940s, Waddington combined the terms "epigenesis" (from embryology) with "genetics" to describe how genetic mutations could influence the development of an organism • This "epigenetics" is today's "developmental genetics" • The valleys indicate cellular states • The ball is a differentiating cell • As the cell differentiates, it makes successive (and irreversible) decisions about which fate to adopt • DNA mutations can change this decision landscape 2nd epigenetics • In 1958, Nanney reinterpreted Waddington's landscapes. • He assumed that "epigenetic" systems are those that underlie cellular memory, i.e. the ways in which a cell "remembers" its differentiation state • In 1970, Riggs and Holliday noticed that DNA methylation might be a mechanism for cellular memory, because it can get passed to the daughter cell in mitosis • This began the view of "epigenetics" as describing the molecular basis of cellular memory 3rd epigenetics: • DNA methylation is just one of many marks distinguishing active from silent genome regions • Others: histone modifications, and other proteins in chromatin • Back-translated to an imagined Greek origin: "epi" ="upon" + "genetics" = "inheritance of variation" • This"epigenetics" is simply "gene regulation" in this course: • Gene expression regulation (today's lecture) • Imprinting • Inheritance of acquired traits?

Impact of population events on genetic variation

• "Bottlenecks" are drastic, temporary reductions in population size. • They reduce variation • They can make some rare alleles more common • "Founder effects" are bottlenecks during migration, when a small subgroup breaks off from the main group • Expansions increase variation, especially rare variation

Nucleosomes are the central unit of chromatin

• "Chromatin" is DNA and its associated proteins • Chromatin serves two functions: 1) package 4 meters of DNA into the nucleus, 2) regulate gene expression • DNA is tightly bound to positively charged histones • DNA accessibility is a major determinant of gene expression (e.g. heterochromatin is typically repressed) • DNA accessibility on nucleosomes is controlled by nucleosome composition and histone modifications

Local vs. distant eQTLs

• "Distant" or "trans" eQTLs are located far from their target gene • They are typically located on different chromosomes. • They act via transregulatory factors • Distant eQTLs can alter the protein sequence of a trans regulator • Or they can alter the abundance of a trans regulator distant eQTLs not well understood • In eQTL mapping, we perform one GWAS analysis for each expressed gene • Therefore, we must correct for even more multiple tests than in normal GWAS: the number of SNPs times the number of expressed genes • At the same time, large samples are hard to get for eQTL studies because they require access to tissues • As a consequence, sample sizes are in the 100s - low, compared to standard GWAS • To avoid the multiple testing burden, most studies test only for local eQTLs around each gene • Our knowledge of distant, trans eQTLs in humans is limited

Components of chromatin and gene regulation

• DNA methylation • Histone marks • Enhancers and 3D chromatin structure • Transcription factors

protein level variation

• For most genes, the protein products perform the actual work in the cell and the body • But measuring protein levels in many samples is hard and expensive • Most research has focused on molecular traits that can be measured by Illumina sequencing • The few studies that did map "pQTLs" (protein QTLs; regions that influence protein levels) showed a surprising amount of discordance between mRNA and protein results

Miller Syndrome

• Multiple malformation disorder affecting face & extremities • Extremely rare: only 30 known cases • Causal gene unknown • Mode of inheritance unknown: potentially autosomal dominant or recessive • No known linkage region

Summary - ways in which traits can be complex

• One strong, "Mendelian" locus plus modifiers (CF) • Some large effects plus many small effects (T1D, eye color, hair color, skin color) • Many small effects (height) • Mendelian alleles in rare families, many common variants of small effect (heart disease) • Extremely many variants of extremely small effect (cognitive traits)

Chromatin in the nucleus

• Repressed regions tend to be associated with the nuclear lamina • Active gene regions tend to be located in the nucleolus and transcription factories • This is where enhancers may loop in to meet and activate promoters

Polygenic risk scores in common disease

• We can use GWAS results (specifically, the effect sizes of associated SNPs) to calculate personal "scores" to quantify an individual's risk of developing a given disease. • These scores are not yet very reliable for most people. • But they can give actionable for those individuals that are most at risk. Polygenic risk scores currently perform worse in non-European populations (SNPs have additive effects)

Common disease-common variant hypothesis

Disease genes from linkage analysis do not explain much of the variation in common diseases. Rare variants with large effect (as identified from pedigrees) do not underlie common genetic disease. Are there many variants at high frequency in the population (i.e. "common"), each with more modest effects? Can we find them by studying populations of unrelated individuals?

Type 1 diabetes

In diabetes, the body does not produce insulin, resulting in abnormal metabolism and elevated glucose levels in the blood • In T1D, an autoimmune reaction destroys the ß-cells of the pancreas, which normally produce insulin. • T1D usually manifests in childhood or adolescence • Incidence: ~0.2% in white populations, lower in African & Asian populations • MZ concordance: 40%, DZ concordance: 5% => large genetic contribution to disease risk • The MHC (major histocompatibility complex) is the most highly polymorphic region in the human genome • It encodes over 200 genes, many with immune functions • Variation in this region is very complex, with nested structural variants, and >2,000 alleles around the world • Many diseases have associations here, especially those involved with (auto)immune disease • While MHC is the most significant association for T1D, there are many additional loci in the genome that contribute

linkage

Linkage is measured by determining the amount of recombination (by meiotic crossing over) between two loci. Frequent recombination means that the two loci are not tightly linked (i.e. they are physically far from each other). Rare or no recombination means that the two loci are tightly linked (i.e. they are close to each other). Linked loci "travel together" through the generations. Linkage is related to genetic distance -> "Genetic" distance is measured in units of recombination events "Physical" distance is measured in base pairs (bp)

Evidence of linkage

Statistical analysis of linked markers with disease phenotype. We're looking for markers with low θ with the disease (i.e. markers that do not often recombine away from the disease). To measure statistical significance, we use the LOD (Log of the ODds): LOD = log10 [L ( linked, θ < 0.5) / L ( not linked, θ = 0.5)] A LOD ≥ 3 is considered significant • The process in the example is repeated for many markers that cover the entire genome • Linkage analysis can be visualized as a "LOD profile" along the genome, where "peaks" indicate the critical intervals

eQTLs

• "Expression quantitative trait loci" (eQTLs) are regions of the genome that contain one or more DNA variants that are associated with the expression level of one or more genes • Just like in GWAS for any other trait, the causal variant is not typically known • Collect RNA levels and genotypes from a group of individuals • For each gene, scan the genome for associations between mRNA levels and genotype • Repeat the process for every gene that is expressed in the tissue

The evolutionary fate of a new mutation depends on its effect on fitness

• "Fitness" depends on the reproductive success of an individual • The number of viable, fertile offspring • Fitness is relative to the population: a fitness of 1 is the same as the population average • Fitness depends on the environment • Fitness is NOT the same as health, strength, size, ... • Some diseases may not reduce fitness! • Fitness is hard to measure in practice • The distribution of fitness effects shows the fraction of mutations with a certain fitness cost • Negative "selection coefficients" reduce fitness • In humans, most new mutations have small or no effects

Local eQTLs

• "Local" or "cis" eQTLs are located close to the gene they influence. • The definition of "close" is arbitrary and differs between studies. It varies from 50 kb to 1 Mb.

statistical power

• "Power" is the probability of detecting a true difference between groups (or association between SNP and trait) • Power is higher for larger sample sizes • Power is higher for larger effects • Power is higher for SNPs with higher allele frequency • When power is low: - true effects are missed - effect estimates are less precise (can be wrong direction!) - some "detected" effects are false positives

humans have low genetic diversity

• 1/1000 base differences between two typical genomes • 5-fold less than mice, 15-fold less than fruit flies

Missing heritability (in height)

• Common variants in the population typically have small effects • Known mutations with large effect are very rare, and behave like Mendelian mutations • There probably are many variants in between • These may account for some of the missing heritability

Mapping DNAse I hypersensitive (DHS) sites to transcription factor binding

• DHS sites mark open chromatin • DHS sites agree well with transcription factor binding sites defined by CHIP-Seq • We can scan DHS sites for DNA motifs and ask which factor might have bound here

Global distribution of genetic variation

• Is most variation specific to a given continent? •Which continent has the most variation? • Outside this continent, where does most variation come from? •Which continent has the most variants not seen anywhere else? human movement and bottleneck events In some populations, globally rare variants are common

The 1,000 genomes project

• Sequence genomes from human populations around the world • 2,504 people • 26 populations • All whole-genome sequenced • 88 million variants discovered • Most identified variants are SNPs and short indels

Sources of gene expression variation

• The heritability of gene expression (i.e., how much of the variation in expression among individuals is due to genetics) varies. The median is about 20%. • Most of the expression heritability (~2/3) is due to trans acting variants

Gene regulation and human disease

• The majority of GWAS hits are in non-coding regions. • They do not change protein sequence. They are also not in LD with other variants that change coding sequence. • To understand how these variants can influence traits and disease, we must understand gene regulation (that is, "epigenetics").

Population genetics

• The science of genetic variation in populations • Where does variation come from? • How does variation change over time? • How is variation distributed among populations? • How does variation influence traits?

Prioritizing potential disease variants

• Variant: • High quality sequencing data • Predicted to disrupt a gene • Not common in the population • Correct segregation pattern in the family (dominant? recessive? de novo mutation?) • Gene: • Gene expressed in "correct" tissue • Gene is not often mutated in the population • Gene is known to affect similar phenotypes (in humans are other species)

reagents for gene mapping

disease phenotype: accurate definition, reliably detectable study population: extended and nuclear family genetic markers: must be polymorphic, must have a known location, must have an assay for genotyping linkage analysis: statistical method that determines random from non random segregation

A variant in the FTO intron influence clinical obesity by disrupting a repressor binding motif, which increases expression of IRX3 and IRX5 in preadipose cells, which results in increased lipid storage in adipocytes

• Epigenetic marks => chromatin states => enhancer annotation • Chromatin conformation capture => hypothesis of effect on distant gene • Gene expression in risk/control carriers => confirm expression effect • Analysis of DNA binding motifs => disrupted repressor motif • Genome engineering of single variant => confirm effects on expression

Hox genes

series of genes that controls the differentiation of cells and tissues in an embryo Body Patterning and Developmental Sequence Unique in that their organization is the same as the tissue organization in the embryo (1st Hox gene is expressed in the head, last is expressed in the tail, have a linear organization in the genome and in the body). This pattern/gradient is in limbs as well Homebox genes: The homeobox was originally described as a conserved DNA motif of about 180 base pairs. The protein domain encoded by the homeobox, the homeodomain, is thus about 60 amino acids long. HOX genes encode a family of transcription factors of fundamental importance for body patterning during embryonic development. Humans, like most vertebrates, have 39 HOX genes organized into four clusters, with major roles in the development of the central nervous system, axial skeleton, gastrointestinal and urogenital tracts, external genitalia, and limbs. HOX gene expression is controlled by polycomb and trithorax complexes, which remodel chromatin and epigenetically silence genes.

signal gradient and signal source

signal is highest at the signal source, decreases as you move away from the signal source to create a gradient The signal levels / thresholds, along with the levels of other signals, determines the cell state/ what genes are expressed

morphogen

A morphogen is a molecule responsible for determining the pattern of tissue development including the positions of the various specialized cell types within a tissue. Morphogens are produced from a localized set of cells and diffuse from the source to form a concentration gradient. Morphogen interact with signaling pathways that produce a specific gene expression profile dependent on the local morphogen concentration. errors in morphogen concentration lead to more of a certain gene expressed, which leads to dysmorphism (example: reduction in overall morphogen concentration leads to less expression of gene above high threshold and more expression of gene below low threshold)

teratogen

A teratogen is an environmental agent which exerts an adverse influence on a developing organism. Different teratogens: Ethanol Smoking Medications Street drugs Maternal Illness Chemical exposure in the workplace Radiation exposure

Gorlin Syndrome or Basal Cell Nevus Syndrome (BCNS) (also Nevoid basal-cell carcinoma syndrome (NBCCS)

BCNS shows autosomal dominant inheritance. About 40% of cases represent new mutations. BCNS is characterized by multiple basal cell carcinomas and diverse developmental defects. 1 case per 50,000-150,000 Patient may have a thousand or more basal cell carcinomas. Only a fraction of 1% become active. Islands of basal cells invade and grow. Bits of precipitated calcium are not uncommon. Numerous palmar pits. Epidermoid inclusion cyst of extremities is frequent. Lamellar calcification of falx cerebri. Bilateral calcifying ovarian fibromas in a 13-year-old female. Odontogenic keratocysts displace teeth and have over 50% recurrence risk. Bifid ribs aid in diagnosis.

Correct action of developmental gene products are dependent on:

Expression of developmental genes in proper location of the embryo, at the appropriate time. Specific amount of protein product present (mutations act as a dominant). Interaction with other developmental gene products.

Six genes can contribute to Waardenburg syndrome

PAX3 (encoding the paired box 3 transcription factor) MITF (microphthalmia-associated transcription factor) EDN3 (endothelin 3) EDNRB (endothelin receptor type B) SOX10 (encoding the Sry box10 transcription factor) SNAI2 (snail homolog 2) Type 1 PAX3 Type 2 MITF (sometimes also SOX10, SNA12, and EDN3/EDNRB) Type 3 PAX3 Type 4 EDN3, endothelin 3 EDNRB, endothelin receptor type B, SOX10

PAX3 Paired Box 3 Transcription Factor

PAX3 encodes a DNA binding transcription factor containing a paired domain. Homologous to the Drosophila paired (prd) gene. PAX3 expression in the mouse (pax3) is present from day 8 to day 17, peaking at days 9 to 12 during neurulation. PAX3 is involved in the development of the central nervous system, somites, skeletal muscle, neural crest-derived lineages including cardiac tissue, melanocytes, and enteric ganglia. It is required to expand a pool of committed melanoblasts or restricted progenitor cells early in development and prevents their terminal differentiation. linked to Type 1 and 3 Waardenburg Syndrome Heterozygous PAX3 mutations are expected to be responsible for most, if not all, WS1 cases. Homozygous or compound heterozygous PAX3 mutations have been described in severe cases of WS3, with extended depigmentation and upper limb defects, sometimes leading to death in early infancy or in utero

Sonic Hedgehog (SHH) - Patched (PTCH1) interaction

PTCH inhibits Smo when SHH is absent (smoothened) SHH present inhibits PTCH, so Smo is active and leads to downstream signaling many components in hh signaling ligand = hedgehog (SHH in vertebrates); receptor = patched; inhibitors/ signal transducers = smoothened, costal-2, and fused; trans-acting factors = cubitus interruptus (Gli 1 2 and 3)

Sonic hedgehog (Shh)

Patched (PTCH1) encodes a receptor for the Sonic Hedgehog gene (SHH). HH genes are conserved throughout evolution and play critical roles in the development and differentiation of vertebrates. The protein made by Hedgehog is thought to boost the activity of several genes including patched. Patched protein builds up until it interrupts transmission of the HH signal from the membrane to the nucleus. By turning off genes that HH turned on, the patched protein keeps the system in check. Sonic Hedgehog (SHH) is a morphogen involved in the development of several organ systems including the brain, spinal cord, eye, craniofacil structures and limbs. SHH is a segment polarity gene - It is secreted at the zone of polarizing activity (ZPA) SHH is required for patterning of the limb - Mutations of SHH result in limb dysmorphology - Left-right body asymmetry Central nervous system development - null mutation cause holoprosencephaly Eye development Mutations in SHH are responsible for causing the most common brain anomaly in humans, holoprosencephaly (HPE). Literally, HPE means "one forebrain" because the two hemispheres of the brain do not form properly in utero. SHH gradient in hand: high concentration = pinky, low = thumb

Types of Waardenburg Syndrome

Type 1 Hypopigmented patches, heterochromia irides, Dystopia canthorum and sensorineural deafness Type 2 Hypopigmented patches, heterochromia irides and sensorineural deafness No dystopia canthorum Type 3 Hypopigmented patches, heterochromia irides, dystopia canthorum, senorineural deafness and musculoskeletal abnormalities Type 4 Shah-Waardenburg syndrome including deafness, depigmentation and intestinal aganglionosis (called Hirschsprung disease [HD])

Where does all this signaling and morphogen asymmetry begin?

at the very beginning of development Asymmetrical deposition of maternal mRNA and proteins Point of entry of sperm Site of previous cytokinesis Contact point with uterine wall

limitations of short read sequencing

repetitive regions transpositions complex rearrangements

a typical human genome in numbers

• 3 - 4 million variants compared to the reference genome • 1.5 - 2.5 million heterozygous sites • 10,000 - 15,000 "singletons" not found in other genomes • Structural variants cover more bases than SNPs: more than 2,000 large variants, which cover 20 million bases • 24 - 30 variants implicated in rare disease • 20 - 35 genes completely knocked out (as compound heterozygotes)

The Genotype-Tissue Expression (GTEx) consortium

• 44 tissues • 449 donors • 7,051 samples • RNA sequencing to 78 million reads / sample Different tissues use the genome very differently Tissues group according to similarities in gene expression Blood, cell lines, and brain are unlike other tissues • Local eQTL mapping • 153k eQTLs for 19,275 genes • 93% of all common variants are at least weakly associated (p < 0.05) with the expression of one or multiple genes in their vicinity. • Some variants are associated with 30 genes in their vicinity. • For 10% of variants, their strongest association varies between tissues. • 52% of GWAS hits also are an eQTL. 26% are eQTLs for multiple genes. • Only 42% of GWAS / eQTL pairs co-localize with their closest gene • This extreme abundance of regulatory variation requires caution when interpreting GWAS hits

The ExAc project

• Aggregates data from >60,000 exomes • De-identified • 7.5 million variants • Very useful as a filter for exome sequencing studies • We can calculate the expected number of mutations that should be seen if the gene were free to mutate without phenotypic consequence • Fewer observed than expected mutations suggests that the gene has important functions • When we do see this gene mutated in a patient, it is more likely to cause disease • The "probability of loss intolerance" (pLI) quantifies how "essential" a gene is some genes can be deleted without apparent consequence • They may not cause disease • Or, their deletion might have too small effects to be "seen" by selection • Or, they may cause disease too late in life to affect fitness

association testing

• Association studies determine whether a genetic variant is "associated with" a phenotype. • We declare an association if the case population has a "significantly" different allele frequency than the control population. • Usually, "p-values" are used to measure statistical significance. The p-value is a probability, with a value ranging from zero to one. We can use it to determine if two populations are "statistically" different. The p-value is the probability of observing a given association (or one that is even stronger than the observed one), if there is no real association. It is a measure of "surprise". • A p-value of less than 0.05 is typically used to accept that an association is statistically significant.

Complexity of "monogenic" diseases - Cystic fibrosis

• CF is a life-threatening disorder that damages the lungs and digestive system through build up of mucus • Caused by mutations in CFTR • Genetic modifiers are alleles at other genes that have an effect on the severity of a Mendelian trait (e.g. pulmonary disease in CF) • Modifiers are a key topic in disease genetics today • They may underlie variation in penetrance of Mendelian traits (sometimes as genetic interactions) • There is discussion if any trait is fully Mendelian, i.e. only due to a single mutation with complete penetrance and not modified by other genes

Consequences of DNA methylation

• DNA methylation is associated with repressed genes • It does so in concert with other factors that result in "repressive" chromatin • DNA methylation may not initiate repression, but lock it in • Many animals (e.g. flies, worms) have no DNA methylation, so clearly DNA methylation is not essential for tight gene repression

enhancers

• Enhancers are distinct genomic regions that contain binding site sequences for transcription factors and that can upregulate ("enhance") the transcription of a target gene from its transcription start site (TSS). • Enhancers can be located far from their target genes, which makes their identification challenging. • Enhancers are often tissue specific • Linking enhancers and their target genes is difficult

The genetics of coronary artery disease

• In spite of a >50% decrease in mortality in recent decades, CAD remains the leading global cause of mortality • More than 900,000 individuals in the US will have a myocardial infarction or die of CAD this year. • Heritable: h2 = 40% - 50% • Well-known environmental & lifestyle risk factors • Gene discovery has taken a path representative of many other complex diseases • GWAS started in 2007 • Some early hits in known risk factors (e.g. LDLR) • Some of the earliest hits are still not understood: e.g. a region on 9p21: perhaps due the ncRNA ANRIL regulating the genes CDNK2A & CDNK2B • 60 loci have been mapped • Most have minor allele frequencies >5%, relative risks < 20%, and together explain 30 - 40% of h2 • Most hits are non-coding • Look for enriched mutations in 5,000 exomes from 5,000 cases • Nine genes identified • Include LDLR (4- fold risk increase) • PCSK9 knock outs were protective • CAD research has enabled genetically informed drug development • E.g. two PCSK9-antibodies that were approved in 2015 had been developed based on observation of PCSK9 knockouts reducing LDL • Cost per approved drug today: $1.4 billion (!) • High cost is mostly due to 95% failure in clinic, either due to low efficacy or unanticipated toxicity • Sometimes, genetic variation has run the tests for us: can see which genes are tolerated if absent, can see side effects based on pleiotropic associations • E.g. a common variant in HMGCR (the target of statins) associated with lower LDL but also with T2D risk - just like statins!

"Genetic drift" is the null model of evolutionary genetics

• In the absence of other factors, a variant will drift • We can use mathematical descriptions of drift to form expectations of how variation in a species is distributed • Deviations from this expectation indicate that other factors might play a role • This is how we "see" the action of evolutionary selection, population history, etc.

GWAS over the years

• Initial studies of a few hundred or thousand people found just a few associations • As sample sizes increased, more and more loci were found • Most of these loci have very small effects • We usually do NOT know the causal variant (because of LD) • Sample sizes are now often larger than 10,000 • GWAS research is highly collaborative • We can now see associations even for extremely complex traits, such as schizophrenia, depression, or intellectual attainment

calling short insertions and deletions

• Insertions and deletions are referred to as "indels" • Short indels (a few bp) can be accurately called from short reads • Longer indels are much harder

liability and risk

• Liability is a mathematical description that allows us to relate binary disease states ("yes or no") to a continuous quantity • It describes a quantity in the organism that results in disease when it is above or below some threshold • Liability can correspond to known physiological measures (e.g. blood pressure might be considered a liability for heart disease) • Or it can be hypothetical, and reflect an unobserved or unknown aspect of physiology • For many common diseases, we do not know the underlying biology!

Chromosome conformation capture (3C)

• Maps physical molecular interactions between genomic elements • Can be combined with capture methods for specific genome regions (e.g. all enhancers for one promoter) • Many varieties: 3C, 4C, 5C, Hi-C Linking human enhancers to their target genes is difficult!

Complex disease and regulatory variation

• Most SNPs identified by GWAS for common diseases are located in noncoding DNA (promoters, enhancers, intronic, intergenic,...) • These SNPs are too far away from coding exons to be in linkage disequilibrium with nonsynonymous variants • They must influence traits by altering gene expression • Often, these effects involve expression changes of genes that are some distance away from the causal variant.

Random walks of variants

• Most new mutations are lost quickly • It takes time for a new mutation to rise to high frequency • Common variants are old • New variants are rare in the population • Deleterious variants are rapidly eliminated by selection. • They are rare in the population • Adaptive variants spread more quickly than neutral variants • Once they are "fixed", they are no longer polymorphic • Therefore, we rarely expect to see adaptive mutations "in the act" of spreading

The human allele frequency distribution

• Most variants in the human population are "rare" (i.e. are present in few people) • But most variants in each genome are "common". Only 1-4% of variants in your genome have ≤0.5% population frequency.

The neutral theory of molecular evolution

• Most variants that exist in a population are evolutionarily "neutral". • Their effects on fitness are too small to be acted on by selection. • In every generation, neutral variants will randomly drift up and down in frequency due to stochastic sampling of gametes. • This process is called "genetic drift".

Obesity and the FTO gene

• Obesity affects 500 million people worldwide, and is a major risk factor for heart disease, type 2 diabetes, and cancer. • Body Mass Index has a heritability of 40-80%, and is a complex trait • GWAS showed that variants in the FTO gene are associated with obesity • The GWAS SNPs are located in an intron of FTO, suggesting that they act by altering gene expression • Mice with FTO knocked out grow smaller, suggesting that FTO is the causal gene • How this happens was unclear. FTO is expressed in the brain. Perhaps FTO regulates appetite? The associated SNPs in FTO are in physical contact with the promoter of the neighboring IRX3 gene Variants in FTO regulate IRX3 and IRX5 • Compare gene expression in primary preadipocyte cells from patients with risk and non-risk variants: • The FTO SNPs were confirmed to have a regulatory effect on IRX3 and IRX5, but not on FTO

PCSK9, cholesterol, and a genetic path to a drug

• PCSK9 encodes an enzyme that binds the LDL receptor • LDL (the "bad" cholesterol) is a major risk factor for heart disease • Some individuals with very low LDL have homozygous nonsense mutations in PCSK91 • Blocking PCSK9 can lower LDL in patients in which statins don't help • FDA approval in 2015

Consequences of genetic diversity for disease gene discovery

• Populations differ in their allele frequencies • Need population specific filters for Mendelian sequencing • Need population specific (or globally applicable) genotyping arrays (currently biased towards Europeans) • Non-European populations carry disease variation not present in other populations • By studying diverse populations, we can discover more disease genes

The ENCODE project

• The "Encyclopedia of DNA Elements" • The project generated systematic maps of transcription, transcription factor binding, chromatin structure, histone modification,... • 1,640 genome-wide data sets • 147 cell types • Data is publicly accessible assays used: • Bisulfite sequencing (DNA methylation) • RNA sequencing (expression and splicing, gene structure, discovery of new genes, chimeric genes in cancer or translocations) • ChIP-seq (DNA which is bound by protein / transcriptions factors) • DNAseI-seq (DNAse I hypersensitive sites mark regulatory DNA regions) • 75% of the genome is transcribed • 10 - 12 splicing isoforms per gene per cell type • Gene expression levels span a million-fold range in abundance • Classifying the genome into distinct chromatin states reveals ~400,000 enhancer and >70,000 promoter-like regions • A repository of genome-wide datasets probing various aspects of genome regulation • Highly useful for relating gene regulation to disease • Still growing: now called "NIH Roadmap Epigenomics Mapping Consortium" (http://www.roadmapepigenomics.org) • A data resource for computational integrationg

Lessons for complex traits from height

• The genetic basis of quantitative traits can involve thousands of loci • Each locus typically has very small effects • Extreme phenotypes (very tall or short; or disease) happen when many alleles with effects in the same direction come together in a person • Some hits tend to occur in the same biological pathway, pointing to the biological basis of the trait • For almost all GWAS hits, we do NOT know the causal variant!

Why is disease gene identification hard?

• The genome has 3 billion bases • ~20,000 protein-coding genes • many additional non-coding genes • hundreds of thousands of regulatory elements • ... and any one of these could be responsible

"Genetic architectures" of complex traits

• The term "genetic architecture" describes how genetic variation influences a trait: • How many loci (genes)? • What are their effect sizes? • Do they interact with each other? • Do they interact with the environment? • What is their molecular basis? • Modifiers of Mendelian disease (cystic fibrosis) • Major loci (type 1 diabetes) • Polygenic architectures (height & heart disease) • Highly polygenic architectures (cognitive & neuropsychiatric disorders)

Biological consequences of the FTO GWAS variants

• The variants alter metabolism directly in fat tissue: risk allele carriers express genes involved in lipid storage more highly, and genes involved in energy production less highly • Risk allele carriers also have larger adipocytes, and fewer mitochondria in their adipocyte cells • Mouse experiments showed that these effects were directly due to different expression of IRX3 and IRX5 The causal variant disrupts a conserved repressor motif • Analysis of evolutionary conservation showed that one of the variants in the haplotype changed a binding motif for the repressor ARID5B • CRISPR-editing confirmed the effects of this causal variant (in FTO) on the expression of IRX3 and IRX5 • The risk allele increases enhancer activity

histone marks

• There is a huge variety of covalent marks that can be attached to histones • There is a large number of enzymes that "write", "read" and "erase" these marks • Many of these enzymes are by themselves not sequence specific. To target specific genome regions, they rely on cofactors, chromatin state, and transcription factors. • Histone marks correlate with gene activity • Whether they are themselves causal for gene state (e.g. expressed or not) is not entirely clear • Either way, they are informative about the state of the given genome region

Thinking about "epigenetic" marks

• These marks are informative about the function of a given genome region • But correlation is not causation • Which mark attracted which other mark? Who comes first? • Many "writers" have no sequence-specificity. Who directs where specifically the marks go? • Transcription factors may guide all this activity, but their binding can be affected by chromatin state as well • Certain "pioneer factors" may be the ultimate master regulators

Recap: Key facts about global human genetic variation

• Two typical human genomes differ at 1/1000 variants • This amount of variation is low compared to other species • Most variation is found in African individuals and among African populations • Non-African populations harbor subsets of African variation because humans spread out of Africa • Most common variants are shared among all human populations • Rare variants are more population specific: "globally rare but locally common"

Roles of DNA methylation

• X-inactivation • Imprinting • Repress germline-specific genes in the soma • Suppress cryptic promoters in gene bodies? • Repress selfish DNA elements • Might DNA methylation permit trans-generational memory? No -> Two waves of demethylation during development