On the estimation of inbreeding depression using different measures of inbreeding from molecular markers

The inbreeding coefficient (F) of individuals can be estimated from molecular marker data, such as SNPs, using measures of homozygosity of individual markers or runs of homozygosity (ROH) across the genome. These different measures of F can then be used to estimate the rate of inbreeding depression (ID) for quantitative traits. Some recent simulation studies have investigated the accuracy of this estimation with contradictory results. Whereas some studies suggest that estimates of inbreeding from ROH account more accurately for ID, others suggest that inbreeding measures from SNP-by-SNP homozygosity giving a large weight to rare alleles are more accurate. Here, we try to give more light on this issue by carrying out a set of computer simulations considering a range of population genetic parameters and population sizes. Our results show that the previous studies are indeed not contradictory. In populations with low effective size, where relationships are more tight and selection is relatively less intense, F measures based on ROH provide very accurate estimates of ID whereas SNP-by-SNP-based F measures with high weight to rare alleles can show substantial upwardly biased estimates of ID. However, in populations of large effective size, with more intense selection and trait allele frequencies expected to be low if they are deleterious for fitness because of purifying selection, average estimates of ID from SNP-by-SNP-based F values become unbiased or slightly downwardly biased and those from ROH-based F values become slightly downwardly biased. The noise attached to all these estimates, nevertheless, can be very high in large-sized populations. We also investigate the relationship between the different F measures and the homozygous mutation load, which has been suggested as a proxy of inbreeding depression.