Abstract

Association mapping of cellular traits such as gene expression can provide powerful insights into the functions of human genetic variation. However, until recently mapping studies have been limited to a very restricted range of tissues or cell states. An additional limitation of association mapping is that, depending on the structure of linkage disequilibrium, the causal variant that drives an association can be extremely difficult to pinpoint. I will discuss work from our group that attempts to address some of these limitations. First I will discuss our work, as part of the Human Induced Pluripotent Stem Cells Initiative (HIPSCI: www.hipsci.org), to develop IPS Cs as model systems for understanding the functions of human genetic variation. I will show how variation due to the cell line donor contributes to heterogeneity at the level of the epigenome, transcriptome and proteome of IPS Cs, and illustrate how individual quantitative trait loci (QTLs) have confounded past studies of human pluripotent stem cells. Second, I will discuss a new statistical method (RASQUAL) that uses allele-specific signals to improve power and fine-mapping accuracy in association analysis of read-based phenotypes. I will show how RASQUAL handles a range of technical and biological biases in allele-specific signals without requiring data filtering or removal, and significantly outperforms all other existing methods. I will finish by discussing our application of RASQUAL to ATAC -seq data in human LCLs to pinpoint likely causal variants in disease, and to uncover long-range interactions between different regulatory regions.