Clade Distillation for Genome-wide Association Studies

Abstract

Testing inferred haplotype genealogies for association with phenotypes has been a longstanding goal in human genetics with several underlying challenges. A key advantage of these methods is the potential to detect association signals caused by allelic heterogeneity -- when multiple causal variants modulate a phenotype -- in both coding and noncoding regions. Recent scalable methods for inferring locus-specific genealogical trees along the genome, or representations thereof, have made substantial progress towards this goal; however, the problem of testing these trees for association with phenotypes has remained unsolved due to the growth in the number of clades with increasing sample size. To address this issue, we introduce several practical improvements to the kalis ancestry inference engine, including a general optimal checkpointing algorithm for decoding hidden Markov models, thereby enabling efficient genome-wide analyses. We then propose LOCATER, a powerful new procedure based on the recently proposed Stable Distillation framework, to test local tree representations for trait association. Although LOCATER is demonstrated here in conjunction with kalis, it may be used for testing output from any ancestry inference engine, regardless of whether such engines return discrete tree structures, relatedness matrices, or some combination of the two at each locus. Using simulated quantitative phenotypes, our results indicate that LOCATER achieves substantial power gains over traditional single marker testing and window-based testing in cases of allelic heterogeneity, while also improving causal region localization relative to single marker tests. These findings suggest that genealogy based association testing will be a fruitful approach for gene discovery, especially for signals driven by multiple ultra-rare variants. Author summaryFor a given set of individuals and at particular location in the genome, there is an underlying genealogical tree relating those individuals. Due to recombination, this tree is not static but rather varies along the genome. For decades investigators have sought to learn and use these trees to identify regions of the genome that impact human traits and disease. In other words, to find trait-associated trees where different clusters of relatives have, for example, high blood pressure. However, since these trees can be so enormous, it is difficult computationally to build them from DNA samples and difficult statistically to find trees with disease clusters: since each tree encodes so many possible clusters, it becomes hard to distinguish signal from noise. Here, we develop a new statistical method, LOCATER, to efficiently aggregate signals across disease clusters within each tree and thereby detect trait-associated trees. LOCATER can work with any ancestry inference method. We show LOCATER is better at detecting these trees than existing methods. We also introduce a suite of broadly applicable algorithms that make our ancestry inference software, kalis, and LOCATER computationally efficient. LOCATER is designed to work with any ancestry inference method.