Do stronger measures of genomic connectedness enhance prediction accuracies across management units?1

Abstract

Genetic connectedness assesses the extent to which estimated breeding values can be fairly compared across management units. Ranking of individuals across units based on best linear unbiased prediction (BLUP) is reliable when there is a sufficient level of connectedness due to a better disentangling of genetic signal from noise. Connectedness arises from genetic relationships among individuals. Although a recent study showed that genomic relatedness strengthens the estimates of connectedness across management units compared with that of pedigree, the relationship between connectedness measures and prediction accuracies only has been explored to a limited extent. In this study, we examined whether increased measures of connectedness led to higher prediction accuracies evaluated by a cross-validation (CV) based on computer simulations. We applied prediction error variance of the difference, coefficient of determination (CD), and BLUP-type prediction models to data simulated under various scenarios. We found that a greater extent of connectedness enhanced accuracy of whole-genome prediction. The impact of genomics was more marked when large numbers of markers were used to infer connectedness and evaluate prediction accuracy. Connectedness across units increased with the proportion of connecting individuals and this increase was associated with improved accuracy of prediction. The use of genomic information resulted in increased estimates of connectedness and improved prediction accuracies compared with those of pedigree-based models when there were enough markers to capture variation due to QTL signals.

GCA: An R package for genetic connectedness analysis using pedigree and genomic data

“…Individual average PEVD: When pairwise PEVD is first computed at the individual level using equation (5), these estimates need to be aggregated and summarized at the unit level. A calculation of summary PEVD can be traced back to Kennedy and Trus [8] as the average of PEVD between individuals across two units.…”

“…A CD metric measures the precision of genetic values and can be interpreted as the square of the correlation between the predicted and the true difference in the genetic values or the ratio of posterior and prior variances of genetic values u [10]. A notable difference between CD and PEVD is that CD penalizes connectedness measurements when across units include individuals that are genetically too similar [4,5]. A pairwise CD between individuals i and j is given by the following equation.…”

“…Computational speed was improved by integrating C++ code into R code using the Rcpp package [15]. The initial versions of the algorithms and the R code were used in previous studies [4][5][6] and were enhanced further for efficiency, usability, and documentation in the current version to facilitate connectedness analysis. The GCA R package provides a comprehensive and effective tool for genetic connectedness analysis and whole genome prediction, which further contributes to the genetic evaluation and prediction.…”

“…A single phenotype with a heritability of 0.6 and a fixed covariate of sex were simulated. This was followed by simulating units using the k-medoid algorithm [18] coupled with the dissimilarity matrix derived from a numerator relationship matrix as shown in previous studies [4][5][6]. The data set was stored as an R object in the package.…”

“…The concept of genetic connectedness was later extended to measure the connectedness level between reference and validation sets in whole-genome prediction. In general, it was observed that increased connectedness led to increased prediction accuracy of genetic values evaluated by a cross-validation [5].…”