Statistical Properties And Performance Of Pairwise Relatedness Estimators Using Turbot (Scophthalmus Maximus L.) Family Data

The statistical properties and performance of four estimators of pairwise relatedness were evaluated in several scenarios using the microsatellite genotype data from a set of large known full-sibships of turbot. All estimators showed a significant negative bias for the four kinships commonly used in these studies (unrelated: UR, half-sibs, full-sibs and parent-offspring), when allele frequencies of the reference population were estimated from the individuals analysed. When these frequencies were obtained from the base population from which all families proceeded, the bias was mostly corrected. The Wang (W) and Li (L) estimators were the least sensitive to this factor, while the Lynch and Ritland (L&R estimator) was the highest one. The error (mean around 0.130) was very similar in all scenarios for W, L and Queller and Goodnight (QG) estimators, while L&R was the highest error-prone estimator. Parent-offspring kinship resulted in the lowest error, when using W, L and QG estimators, while UR resulted in the lowest error with the L&R estimator. Globally, W was the best-performing estimator, although L&R could perform better in specific sampling scenarios. In summary, pairwise estimators represent useful tools for kinship classification in aquaculture broodstock management by applying appropriate thresholds depending on the goals of the analysis.