SNP-Based Linkage Analysis in Extended Pedigrees: Comparison between Two Alternative Approaches

Abstract

Background: Linkage analysis on extended pedigrees is often challenged by the high computational demand of exact identity-by-descent (IBD) matrix reconstruction. When such an analysis becomes not feasible, two alternative solutions are contrasted: a full pedigree analysis based on approximate IBD estimation versus a pedigree splitting followed by exact IBD estimation. A multiple splitting (MS) approach, which combines linkage results across different splitting configurations, has been proposed to increase the power of single-split solutions. Methods: To assess whether MS can achieve a comparable power to a full pedigree analysis, we compared the power of linkage on a very large pedigree in both simulated and real-case scenarios, using variance components linkage analysis of a dense SNP array. Results: Our results confirm that the power to detect linkage is affected by the pedigree size. The MS approach showed higher power than the single-split analysis, but it was substantially less powerful than the full pedigree approach in both scenarios, at any level of significance and variance explained by a quantitative trait locus. Conclusion: The MS approach should always be preferred to analyses based on a single split but, when adequate computational resources are available, a full pedigree analysis is better than the MS analysis. Rather than focusing on how to best split a pedigree, it might be more valuable to identify computational solutions that can make the IBD estimation of dense-marker maps practically feasible, thus allowing a full pedigree analysis.

PBAP: a pipeline for file processing and quality control of pedigree data with dense genetic markers

“…Smaller pedigrees can be analyzed with exact computation, while larger pedigrees require Markov chain Monte Carlo (MCMC) sampling (Lange and Sobel, 1991;Thompson, 1994). It is important to note that full pedigree analysis based on IBD probability estimation can be substantially more powerful than approaches that split the pedigrees followed by exact IBD computation, thus emphasizing the need for use of MCMC sampling methods in the context of large pedigrees (Saint-Pierre et al, 2014).…”