L33 Genomic Medicine Flashcards
ExAC
see onenote slides and read own notes
Exome Aggregation Consortium
- ~60,000 mostly healthy human exomes
- exome = full sequence of protein coding region of a genome
ExAC gives us a view of the coding mutational load in humans far more precise than anything before
ExAC is transformative for clinical applications
see onenote slides
- before ExAC, our knowledge of mutation pathogenicity was limited to small sampled sizes
- all of us carry rare variants
- differentiating the rare variants that cause Mendelian disease from the ones that are simply there is hard, because they are all rare
Rare prion disease in ExAC
see onenote
fatal familial insomnia = rare prion disease caused by mutations in the gene PRNP
- Did exome sequencing and found she (Sonia) had mutation in PRNP gene => fatal familial insomnia
- In ExAC, there were 52 people with variants in the PRNP gene but they got an expectation of only 1.7 people to get fatal familial insomnia => not all alleles had equal penetrance
Reclassifying rare variants
see onenote
- on the basis of ExAC data, able to confirm that Sonia is a carrier of a pathogenic allele and will get FFI
- Sonia had 100% pathogenic variant
- Plotted freq of variant (y axis) vs those reported to cause a prion disease (x axis)
- Depending on where the mutation occurs in the gene, it determines whether the mutation is pathogenic or not
How to deliver on the promise of genomic medicine
see onenote diagram
UK biobank
~700,000 individuals, 2000+ phenotypes from each of them
- massive population cohort that will be re-phenotyped multiple times in the future to gather data about late onset disease
- genotyped with SNP-chip (tag-SNP based)
Predicting disease with polygenic risk scores
- GWAS hits can be informative even if the underlying biology is unknown
- take every single SNP ever associated with a trait and count the number of risk alleles you see across all of them in an individual
Polygenic risk scores (PRS)
- Control tend to have lower polygenic risk scores
- Combine PRS and clinical risk estimates to recommend specific interventions for each patient
- Genetic risk vs clinical risk => help be more nuanced about intervention for each patient
- PRS supersedes family history
- PRS works better in individuals from the same population as the one where the data came from
What if we do want to understand the biology
see onenote
The vast majority of trait-associated loci are non-coding
- BUT they clearly do something
- So they must be doing it through perturbing the gene’s expression somehow
QTL = genetic region associated with a particular quantitative phenotype
cis vs trans QTLs
differ in mechanism of action
- Cis = SNP in the same chromosome as the gene it affects, allele-specific manner e.g. affect chromosome with C allele but not the other chromosome with A allele
- Trans = SNP in different chromosome as the gene it affects, SNP probably codes for a TF, diffusible, affects both copies of the gene
- Is every cis QTL some other gene’s trans QTL?
When do QTLs matter?
- response eQTL
see onenote slides
- Response eQTL are only eQTL in a particular environment e.g. under different stimuli in the same tissue, often seen in immune reponses or in healthy vs disease states
- Response to doxorubicin (cancer drug) varies depending on genotype, can cause cardiotoxicity; being able to predict cardiotoxicity before commencing treatment would help patient outcomes
- Since we’re always studying healthy people, are we missing a lot of the variation out there
- Variable penetrance
- associations between cis-QTLs and coding variants provide a possible mechanism for this observation
From individual genes to networks
- genes do not act in isolation from one another
- organise our knowledge of genes into networks, more representative of a cell
- Protein-protein interaction
- Regulatory network, there is direction e.g. gene A (TF) impacts expression of gene B
- Most human traits are polygenic
Gene co-expression network
- adds to our understanding of trait architecture
- When gene A go up, does gene B go up? Etc.
- Do it pair by pair
- Leads to understanding of clumps of genes that work together
- Guilt-by-association prediction, new way of discovering targets and network of genes that underlie a trait
Endless small effects?
- if you disregard significance, every SNP test by GIANT seems to contribute to human height
- similar results in other traits
- does this mean all SNPs contribute to all traits?
The same effects across all traits?
The same effects across all traits
- GWAS usually flag the same loci across similar disorders, suggesting there is a base of common genetic architecture
- but disease do differ, so how is that specificity encoded?
- GWAS results suggest that most human traits are very much polygenic
