Groupwise Association Test for Rare Variants

We developed methods to detect associations of rare variants under low-coverage sequencing or pooling. The paper will be online soon, and here is the reference. To download software, please visit Low Coverage & Pool sequencing.

Oron Navon*, Jae Hoon Sul*, Buhm Han, Lucia Conde, Paige Bracci, Jacques Riby, Christine F. Skibola, Eleazar Eskin, Eran Halperin. Rare Variant Association Testing Under Low-Coverage Sequencing. Genetics. 2013 May 1. [Epub ahead of print] PubMed PMID: 23636738.

We developed two methods to detect associations of groups of rare variants. Those two methods are RWAS (Rare variant Weighted Aggregate Statistic), and LRT (Likelihood Ratio Test). They work under the assumption that genotypes are generated with high confidence.

RWAS (Rare variant Weighted Aggregate Statistic) is a groupwise association test for identifying associations of groups of rare variants. RWAS groups variants and computes a weighted sum of differences in mutation counts between case and control individuals. Weights of RWAS are estimated from data to achieve nearly optimal power under a disease model in which all variants make an equally small contribution to population disease risk. For more information on RWAS, please refer to the following paper

Jae Hoon Sul, Buhm Han, Dan He, Eleazar Eskin. “An optimal weighted aggregated association test for identification of rare variants involved in common diseases.” Genetics (In Press)

LRT (Likelihood Ratio Test) is a method that tries to identify which variants are causal by taking advantage of both prior information (of how likely each variant is functional) and data. LRT uses this information (of which variants are likely causal) to better detect associations of groups of rare variants. To identify causal variants, LRT assumes that some variants are causal and some are not (called “causal statuses of variants”) and computes the likelihood of the data under every possible causal statuses. This allows LRT to compute likelihoods of null and alternative models where the null model is one that asserts no causal variants in a group while the alternative model asserts at least one causal variant. A statistic of LRT is a ratio between likelihoods of the two models, and the permutation test is performed to obtain the significance of the statistic. For more information on LRT, please refer to the following paper

Jae Hoon Sul, Buhm Han, Eleazar Eskin. “Increasing Power of Groupwise Association Test with Likelihood Ratio Test.” In Proceedings of the Fifteenth Annual Conference on Research in Computational Biology (RECOMB-2011). Vancouver, Canada: March 28th-31st, 2011

Jae Hoon Sul, Dan He, Buhm Han and Eleazar Eskin are supported by National Science Foundation grants 0513612, 0731455, 0729049 and 0916676, and NIH grants K25-HL080079 and U01-DA024417. Buhm Han is supported by the Samsung Scholarship. This research was supported in part by the University of California, Los Angeles subcontract of contract N01-ES-45530 from the National Toxicology Program and National Institute of Environmental Health Sciences to Perlegen Sciences.

Contact: Jae Hoon Sul (jhsul at cs dot ucla dot edu)