Volume: 5 Issue 4 [April 2022]

Article:Methodological Comparison of Genomic Selection for Several Traits in Arabidopsis and Soybean

Author: Medrine Mmayi Odinga, Jian-Ying Feng, Yuan-li Ni, Xiu-li Yue, Yuan-Ming Zhang

Volume: Vol 5, Issue 4; April 2022

DOI: North American Academic Research, 5(4) April February 2022, https://doi.org/10.5281/zenodo.6539949

Abstract: Evidence from previous studies has confirmed the advantages of genomic selection (GS) in animal and plant breeding. Although many GS approaches are available, the knowledge about which is the best approach is limited. In this study, we evaluated six parametric and two non-parametric GS methods using the Arabidopsis datasets and soybean datasets with salt and water treatment in 2014 and 2015. Cross-validation experiments, along with all markers and related markers at the 0.05, 0.10 and 0.20 levels of significance in each of the datasets were carried out. Prediction accuracies were obtained, indicated by R2 and Mean Absolute Error (MAE) between actual and predicted values, for all the methods in each dataset. As a result, the average R2 across various traits and different numbers of related markers in Arabidopsis thaliana datasets for BayesA, BayesB, BayesC, BRR, BLASSO, GBLUP, RKHS, and BLUP were between 0.644 and 0.668, 0.700 suggesting that GBLUP and BLUP are the best GS approaches. The same trend was replicated in all the other datasets. EMMA model was used to estimate polygenic and residual variances. The average heritabilities for all markers and part of markers correlated at the 0.05 significance level were 0.758 and 0.741, 0.434 and 0.511 respectively, for Arabidopsis thaliana, soybean 2014 and 2015 water-treatment datasets. The same trend was repeated in salt treatment datasets. This indicates that using markers related to the trait can also capture the genetic variation in the natural population. In summary GBLUP, RKHS and BLUP are the superior GS approaches.

Comments