Our research focuses on the development of statistical methods for uncovering the genetic basis of human disease, and on the population genetics underlying these methods. Areas of interest include:
Disease mapping in structured populations
Structured populations pose several technical challenges for disease mapping. Population stratification motivates the use of mixed linear models (MLM) and/or principal components analysis (PCA), but the question of how to maximize power using these methods is complex. In addition, chromosomal segments of distinct local ancestry in individuals from admixed populations should be accounted for, because of the advantages of incorporating admixture association signals, because different LD patterns in ancestral populations can lead to different effect sizes at SNPs in LD with a causal SNP, and because this information can be useful for fine-mapping. Our ongoing research spans all of these technical challenges.
MLM Perspective paper (Yang et al. 2014). Also see African-American selection paper (Bhatia et al. 2014), Fst paper (Bhatia et al. 2013), PCA SNP weights paper (Chen et al. 2013), MIXSCORE paper (Pasaniuc et al. 2011), Africa selection paper (Bhatia et al. 2011) and Admixture review paper (Seldin et al. 2011).
Heritability of complex traits
Genome-wide association studies (GWAS) have identified hundreds of robust associations, yet have explained only a small fraction of the genetic heritability of human traits. The vast majority of the research that has been conducted thus far has focused on the search for specific disease risk variants, under the view that identifying specific variants is the way to understand genetic risk. Though this paradigm is unquestionably important, its inability thus far to explain the bulk of genetic heritability provides a strong motivation to delve deeper into the underlying properties of the unexplained heritability. Our ongoing research uses both identity-by-descent (IBD) and identity-by-state (IBS) to investigate these questions in related and/or unrelated individuals.
Admixture-based heritability paper (Zaitlen et al. 2014). Also see Local heritability paper (Gusev et al. 2013), deCODE heritability paper (Zaitlen et al. 2013) (Research Highlight) and Gene expression heritability paper (Price et al. 2011).
Functional components of heritability
Functional annotation data represents a great opportunity for the field of human genetics, as there is increasing evidence that genetic associations are concentrated in annotation classes spanning a small fraction of the genome. In contrast to approaches that rely on genome-wide significant associations to identify enriched regulatory and cell-type specific annotation classes, our research takes a heritability approach, leveraging genome-wide polygenic signals.
Functional heritability paper (Gusev et al. 2014)
Statistical methods for sequencing studies
Sequencing provides an increasingly appealing alternative to the traditional GWAS approach of identifying common risk variants using genotyping arrays. Exome sequencing is of particular interest, but will eventually give way to whole-genome sequencing as costs decrease. New statistical methods are needed to analyze the avalanche of data that is currently being generated. The question of how to allocate resources to genotyping arrays, exome sequencing, and/or whole-genome sequencing at various coverage levels is a critical but complex question in the face of rapidly changing costs.