Effects of models with finite loci, selection, dominance, epistasis and linkage on inbreeding coefficients based on pedigree and genotypic information

Considering the effects of selection, dominance, epistasis and linkage, a stochastic computer simulation was performed to study how well inbreeding coefficients calculated from pedigree (f_ped) and genotypic frequencies (f_het) correspond to the inbreeding coefficient that is defined as the proportion of autozygous loci in the modelled genome (i.e. the level of autozygosity, f_aut). Although in random mating populations all three inbreeding coefficients show almost (with slight deviations in models with two loci) the same expectation, they represent three separate variables. First, f_aut, f_ped and f_het responded differently to selection, dominance, epistasis and linkage. Second, they did not have the same standard deviations, which means that the effects of random drift, especially in models under selection, were not affecting all three coefficients in the same way. Finally, they were not always defined in the same domain. With selection as the most important factor responsible for the observed discrepancies, the bias (discrepancy) was present in both directions, thus leading to overestimation or underestimation of the observed level of autozygosity depending on the genetic model, linkage and initial gene frequency. Variation of the autozygosity level (between replicates) was increased notably in models with additive inheritance under selection and was an additional potential source of bias. Thus, when the trait is, to a large extent, controlled by a finite number of loci and when selection is present, the bias in the estimation of the autozygosity is likely to occur and caution is necessary whenever conclusions are based on inbreeding coefficients estimated from the pedigree or decrease in heterozygosity.