CMStatistics 2018: Start Registration
View Submission - CMStatistics
Title: Quantifying genomic connectedness and whole-genome prediction accuracy using bootstrap aggregation sampling Authors:  Gota Morota - Virginia Polytechnic Institute and State University (United States) [presenting]
Abstract: Prediction of complex traits has been a focus of quantitative genetics since the beginning of the 20th century. The advancement of whole-genome prediction has sparked a renewed interest in this topic. In statistics, connectedness is a measure germane to estimable comparisons. In the whole-genome prediction era, the concept of genetic connectedness can be extended to measure a connectedness level between training and testing sets. Recent studies have shown that connectedness across sets increased with the increasing proportion of related individuals, and this increase was associated with the improved accuracy of prediction. However, prior studies on comparing the degree of connectivity mainly used model-based formulas of prediction error variance computed from best linear unbiased prediction, leaving an open question about the possibility of computing empirical connectedness. Therefore, we derived prediction error variance by using bootstrap aggregation sampling and investigated the relationship between empirical connectedness measures and prediction accuracy in the cross-validation framework. We also demonstrated the potential of non-parametric relationship matrices to quantify genomic connectedness and prediction accuracy in the presence of non-additive gene actions.