Comparison of identity by descent estimates with Plink and refinedIBD in dogs

With the availability of dense SNP genotype data various types of estimation methods were developed to estimate relatedness of any two individuals, even in absence of traditional pedigrees. One of the most prominent method was the identity by descent (IBD), widely used in genetic diversity studies. IBD itself could be estimate using different approaches and software that might provide different results. The purpose of this study was to compare the estimates from two established software, probabilistic approach by Plink and a non-probabilistic approach based on haplotypes by refinedIBD. High density SNP genotypes from 98 Leonberger dogs were used to estimate IBD coefficients based on two data types: with one of the SNP markers in high linkage disequilibrium removed, as required by Plink, and SNP markers subjected only to standard quality control, as required by refinedIBD. The Pearson correlation coefficients from pairwise estimates were 0.97 when estimated with the same software and 0.84 between the two software and data types, as required by the respective user manuals. The numerical differences were clustered around zero (i.e. no to little difference) for half of the pairwise comparisons, and up to ±0.1 for the vast majority of cases. The most extreme differences were consistently estimated higher by Plink. Because of these differences a follow up investigation should be done, including pedigrees, as well as simulated data to provide a comprehensive analysis.