An assessment of the reliability of quantitative genetics estimates in study systems with high rate of extra-pair reproduction and low recruitment

Abstract

Quantitative genetics approaches, and particularly animal models, are widely used to assess the genetic (co)variance of key fitness related traits and infer adaptive potential of wild populations. Despite the importance of precision and accuracy of genetic variance estimates and their potential sensitivity to various ecological and population specific factors, their reliability is rarely tested explicitly. Here, we used simulations and empirical data collected from an 11-year study on tree swallow (Tachycineta bicolor), a species showing a high rate of extra-pair paternity and a low recruitment rate, to assess the importance of identity errors, structure and size of the pedigree on quantitative genetic estimates in our dataset. Our simulations revealed an important lack of precision in heritability and genetic-correlation estimates for most traits, a low power to detect significant effects and important identifiability problems. We also observed a large bias in heritability estimates when using the social pedigree instead of the genetic one (deflated heritabilities) or when not accounting for an important cause of resemblance among individuals (for example, permanent environment or brood effect) in model parameterizations for some traits (inflated heritabilities). We discuss the causes underlying the low reliability observed here and why they are also likely to occur in other study systems. Altogether, our results re-emphasize the difficulties of generalizing quantitative genetic estimates reliably from one study system to another and the importance of reporting simulation analyses to evaluate these important issues.

Can dominance genetic variance be ignored in evolutionary quantitative genetic analyses of wild populations?

“…The script of the simulations is provided in Text S2. To assess the identifiability of all random effects, we (i) estimated the correlations between each pair of variance components estimated by the “full model” across all simulated datasets (as done in Bourret and Garant 2017) and (ii) extracted each empirical model's average information matrix and performed a multivariate sampling approach (with n = 1000) advocated in Houle and Meyer (2015) to estimate sampling correlations between all variance components and their 95% CI. A strong negative correlation is expected when two components are confounded (low identifiability).…”

Evolutionary potential of morphological traits across different life‐history stages

“…The pedigree was built based on genetic information following Bourret & Garant (in press). Briefly, dam identities were obtained from field observations (i.e.…”

“…Here, we focus on morphological traits measured in nestlings during their development until fledging. Nestlings are easily captured to take measurements at different ages, and, given the natural half‐sib design created by the high EPP rate in this species, we can obtain precise and accurate quantitative genetic estimates for these traits (Bourret & Garant, in press).…”

Effects of blood parasite infection and innate immune genetic diversity on mating patterns in a passerine bird breeding in contrasted habitats

“…Mating patterns were defined for both social and extra-pair couples (males assigned using microsatellite loci—see Bourret & Garant (2017) for details on parentage assignment procedures). In brief, candidate genetic fathers of a given nestling included all males captured on the same farm on both previous and following years and within a 15-km radius of the nestling’s nest box (covers males from one to nine farms, mean = 4.8 ± 1.9 farms (see Lessard et al, 2014 for a justification of this scale).…”