A hidden Markov model to estimate inbreeding from whole genome sequence data

Inbreeding results from the mating of related individuals and has negative consequence because it brings together deleterious variants in one individual. Inbreeding is associated with recessive diseases and reduced production or fitness. Until recently inbreeding was estimated with genealogical data, which have some limitations. Genomic estimates of
inbreeding can now be obtained thanks to new technologies. Methods based on genomic relationships assume either that identical-by-state markers are identical-by-descent (IBD) or correct for the allele frequencies estimated in a « founder generation ». Similarly, with runs of homozygosity (ROH), the hypothesis is that stretches of homozygous markers longer than a determined threshold (measured in number of markers or in Mb) are IBD. Although sequencing data are more informative, most of the recent studies used genotyping data. Due to high genotyping errors rates with sequencing technologies, ROH are difficult to use on such data. We herein describe an original method to model inbreeding along chromosomes. It relies on a hidden Markov model that determines locally whether a segment is inbred and gives a measure of inbreeding's age. The method uses a function modeling the probability to observe a genotype in an inbred segments. It is well suited for sequencing data because the function incorporates genotyping errors and uncertainty associated with low or moderate coverage. We first determine the properties of our new method with different simulation scenarii. Then, we use it to estimate inbreeding in the Belgian Blue Beef cattle population with either genotyping arrays (of various densities) or whole genome sequencing data. The estimates will be compared to those obtained with other methods (pedigree, ROH, genomic relationship). Latest results will be presented.