2013 Poster Sessions : An Accurate Method for Inferring Relatedness in Large Datasets of Unphased Genotypes via an Embedded Likelihood-Ratio Test

Student Name : Jesse Rodriguez
Advisor : Serafim Batzoglou
Research Areas: Artificial Intelligence
Studies that map disease genes rely on accurate annotations that indicate whether individuals in the studied cohorts are related to each other or not. For example, in genome-wide association studies, the cohort members are assumed to be unrelated to one another. Investigators can correct for individuals in a cohort with previously-unknown shared familial descent by detecting genomic segments that are shared between them, which are considered to be identical by descent (IBD). Alternatively, elevated frequencies of IBD segments near a particular locus among affected individuals can be indicative of a disease-associated gene. As genotyping studies grow to use increasingly large sample sizes and metaanalyses begin to include many data sets, accurate and efficient detection of hidden relatedness becomes a challenge. To enable disease-mapping studies of increasingly large cohorts, a fast and accurate method to detect IBD segments is required. We present PARENTE, a novel method for detecting related pairs of individuals and shared haplotypic segments within these pairs. PARENTE is a computationally-efficient method based on an embedded likelihood ratio test. As demonstrated by the results of our simulations, our method exhibits better accuracy than the current state of the art, and can be used for the analysis of large genotyped cohorts. PARENTE’s higher accuracy becomes even more significant in more challenging scenarios, such as detecting shorter IBD segments or when an extremely low false-positive rate is required.

Jesse Rodriguez is a PhD student in the Biomedical Informatics program who works on algorithms for problems in population genetics: relative-finding and ancestry inference.