AB092. Regional IBD analysis (RIA): a new method for linkage analysis in extended pedigrees using genome-wide SNP data

Background: Recent shift of focus in genetic studies to rare variants has revived interest in linkage analysis, which can ascertain effects of a locus regardless of allelic heterogeneity. However, exact calculations for traditional linkage analysis are computationally impractical in large, extended pedigrees, which are often encountered in studies of complex, low-penetrance diseases. Although simulation-based methods can be used in such circumstances, they require significant computational work and are not exact. We propose regional IBD analysis (RIA), a non-parametric linkage method based on comparison of locally and globally estimated identity by descent (IBD) sharing in affected relative pairs (ARPs) for use in these circumstances.

Methods: In RIA, genome-wide SNP data are used to calculate the “global” expected IBD sharing probabilities specific to each ARP, against which a “local” set of IBD sharing probabilities, estimated using SNP data within a window of pre-specified width, can be compared. These IBD sharing probabilities can be estimated using a variety of programs. We used PLINK and KING in this study. The global and local IBD sharing probabilities can be used to construct a non-parametric maximum likelihood statistic (MLS)-like test of linkage in each window. We illustrate the use of our method to detect linkage signals in real nuclear-family data from a study of primary vesicoureteral reflux and in simulated data based on large extended pedigrees from a Brazilian study of visceral leishmaniasis, and compared the results with those obtained from “traditional” methods including the Kong and Cox exponential model LOD score from Merlin and the MCMC-based lm_ibdtests from MORGAN.

Results: RIA successfully detected the linkage signals in these data sets, but with significant reduction in computational time (e.g., 2 vs. 66 h on a dedicated server on a data set consisting of 3,626 individuals from 308 extended families, 357 of whom were affected, genotyped at 545,433 SNPs) and resources compared with the traditional methods.

Conclusions: The proposed method should be useful in studies involving large extended families, in which traditional linkage analysis is not feasible. Additionally, because it does not rely on any prior knowledge about familial relatedness, the method has an additional advantage of being robust to pedigree misspecification and can be used even in absence of pedigree information. RIA is available at www.staff.ncl.ac.uk/richard.howey/ria/.

Keywords: Regional IBD analysis (RIA); non-parametric linkage analysis; identity by descent (IBD); affected-relative-pair method; extended family