A model‐based approach to characterize individual inbreeding at both global and local genomic scales

Inbreeding results from the mating of related individuals and may be associated with reduced fitness because it brings together deleterious variants in one individual. In general, inbreeding is estimated with respect to an arbitrary base population consisting of ancestors that are assumed unrelated. We herein propose a model-based approach to estimate and characterize individual inbreeding at both global and local genomic scales by assuming the individual genome is a mosaic of homozygous-by-descent (HBD) and non-HBD segments. The HBD segments may originate from ancestors tracing back to different periods in the past defining distinct age-related classes. The lengths of the HBD segments are exponentially distributed with class-specific parameters reflecting that inbreeding of older origin generates on average shorter stretches of observed homozygous markers. The model is implemented in a hidden Markov model framework that uses marker allele frequencies, genetic distances, genotyping error rates and the sequences of observed genotypes. Note that genotyping errors, low-fold sequencing or genotype-by-sequencing data are easily accommodated under this framework. Based on simulations under the inference model, we show that the genomewide inbreeding coefficients and the parameters of the model are accurately estimated. In addition, when several inbreeding classes are simulated, the model captures them if their ages are sufficiently different. Complementary analyses, either on data sets simulated under more realistic models or on human, dog and sheep real data, illustrate the range of applications of the approach and how it can reveal recent demographic histories among populations (e.g., very recent bottlenecks or founder effects). The method also allows to clearly identify individuals resulting from extreme consanguineous matings.