Determining the causal links and clinical significance of rare genetic variants

GWAS have identified many common genetic variants associated with various traits and diseases. However most of the individual effects of common variants at the trait level are very small, requiring very large sample sizes (i.e. 10's of thousands) for detection, with associations found providing low accuracy predictions of an individual's liability to disease or outcomes after treatment. Genetic effects of common variants on intermediate phenotypes, such as gene expression or protein concentrations, are often much larger than those on traits and diseases and such associations are thus often detectable in smaller samples of hundreds or thousands of individuals. Combining information from large studies of disease outcomes and smaller transcriptomic or proteomic studies in a two-sample Mendelian randomisation study can be used to provide evidence for a causal path from DNA variation through gene expression to a disease outcome. Even so, the relatively small genetic effects of common variants on phenotypic traits make them difficult to study in the small functional studies feasible in the laboratory.
Some genetic variants have larger genetic effects than those common variants detected by GWAS, but such variants are often kept at low frequency within individual pedigrees by natural selection as a consequence of their larger effects on individual fitness. Such variants are hence difficult to detect in cosmopolitan studies of unrelated individuals, but may become detectable in studies of pedigreed populations, especially where small founder population size and drift may enhance the frequency of otherwise rare variants. Nonetheless such variants are unlikely to be in LD with and hence captured by associations with SNPs on standard arrays. Rare variants of large effect are most likely to be located within or close to expressed genes. Hence using DNA sequence from the exome and adjacent regions is a good strategy to capture such variants.
In this project we propose to link proteomic data with the exome variants to detect locally (i.e. cis) acting genetic effects on protein concentrations.

Large studies of disease outcomes combined with studies of intermediate phenotypes (transcriptome, proteome) in two-sample Mendelian randomisation (MR) can be used to provide evidence for a causal path from DNA variation through gene expression to disease outcome. Even so, their small phenotypic effect size makes common genetic variants detected by GWAS in cosmopolitan populations difficult to follow up functionally. Genetic variants of larger effects size are rare in populations of largely unrelated individuals, but may become detectable in isolate populations with higher levels of kinship. Such variants are unlikely to be captured by standard SNP genotyping and so require more genome sequence-level analysis.
This project will identify novel and rare variants in whole-exome sequence data that have cis-acting effects on protein abundance in 8000 individuals from isolate populations from the Northern Isles of Scotland and from Croatia. Evidence for the role of protein expression in disease aetiology will be explored using two sample MR. These studies will include re-contact of cohort participants carrying rare variants with large predicted phenotypic effects for a more detailed phenotypic assessment and generation of tractable biological samples. Further study will include exploiting proteomic and other 'omic data to build graphical predictive models for individual traits or diseases incorporating multiple loci.