The Jackpot Effect - When Do Family Samples Provide More Power To Detect Trait-associated Rare Variants? S. Feng1, G. Pistis1,2, A. Mulas2, M. Zoledziewska2, F. Busonero2, S. Sanna2, D. Liu3, F. Cucca2, G. R. Abecasis1 1) Biostatistics, University of Michigan, Ann Arbor, MI 48109; 2) Istituto di Ricerca Genetica e Biomedica-CNR, Monserrato (CA), 09042, Italy; 3) Department of Public Health Sciences, College of Medicine, Penn State University, Hershey, PA, 17033.

   Population samples have been used to discover many trait associated common variants using array-based GWAS, but may not be the optimal choice for detecting rare variants with moderate to large effects. Here, we set out to compare the power of population and family samples to study trait associated rare variants. We describe settings where family samples can provide more power than population samples for rare variant association studies. In particular, we show that in population samples observing enough trait associated alleles in each causal gene can be extremely challenging and require very large samples. In contrast, in family samples that include large numbers of related individuals, there will more often be one or more genes where many observations of a trait associated allele are made.
   For example, by simulation, we show that when there are 100 loci where trait associated variants segregate at a frequency of ~0.1% and modify a quantitative trait by one standard deviation, a study of 5,000 unrelated individuals provides ~60% power to detect at least one of these loci at genomewide significance. In contrast, a study of 5,000 individuals distributed across 100 large families, each with ~50 individuals, provides ~90% power to detect at least one of the trait associated loci in the same setting. For gene-level association tests similar results are observed and the contrast is even more dramatic when association signal is dominated by singleton variants, defined as variants present in a single founder or unrelated individual.
   The advantages of family samples are due to a Jackpot effect, where multiple copies of some trait-associated rare alleles are shared among individuals with a common ancestor, greatly increasing power. We show that the advantages of family samples for rare variant studies increase as the number of individuals with a shared recent ancestor increases. Using simulations based on an isolated population sample from the island of Sardinia, we show often dramatic differences in power. Our results provide guidance to investigators hoping to identify trait associated rare variants and deciding between family and population based designs.

You may contact the first author (during and after the meeting) at