What should we weigh to estimate heterozygosity, alleles or loci?

The interest to study the effects of inbreeding in natural populations has increased in the last years. Several microsatellite-derived metrics have recently been developed to infer inbreeding from multilocus heterozygosity data without requiring detailed pedigrees that are difficult to obtain in open populations. Internal relatedness (IR) is currently the most widespread used index and its main attribute is that allele frequency is incorporated into the measure. However, IR underestimates heterozygosity of individuals carrying rare alleles. For example, descendants of immigrants paired with natives (normally more outbred) bearing novel or rare alleles would be considered more homozygous than descendants of native parents. Thus, the analogy between homozygosity and inbreeding that generally is carried out would have no logic in those cases. We propose an alternative index, homozygosity by loci (HL) that avoids such problems by weighing the contribution of each locus to the homozygosity index depending on their allelic variability. Under a wide range of simulated scenarios, we found that our index (HL) correlated better than both IR and uncorrected homozygosity (H(O)), measured as proportion of homozygous loci) with genome-wide homozygosity and inbreeding coefficients in open populations. In these populations, which are likely to prevail in nature, the use of HL instead of IR reduced considerably the sample sizes required to achieve a given statistical power. This is likely to have important consequences on the ability to detect heterozygosity fitness correlations assuming the relationship between genome-wide heterozygosity and fitness traits.