181
Genomic Prediction and Genome Wide Association in Humans with Whole Genome Sequence Data
The transition from GWAS chip to sequencing data with increasingly larger sample sizes has many ramifications for efforts to conduct genomic prediction and genome wide association studies. First, as data sets grow larger, it is of interest to consider methods whose running time is linear in the data size. Second, it can be beneficial to model non-infinitesimal genetic architectures whose distribution of effect sizes is different from Gaussian, including minor allele frequency (MAF) dependent architectures. Third, although the fact that mixed model association can be viewed as a test for association on phenotypic residuals of BLUP predictions motivates a generalization to phenotypic residuals of predictions based on non-infinitesimal genetic architectures, this will require new approaches to calibration of test statistics. In this invited talk, we review recently published work in all of these research directions.
Keywords:
Prediction
Genome wide association
Sequence data
