Abstract
- GS
- Genomic Selection
- BLUP
- Best Linear Unbiased Prediction
- EBVs
- Estimated Breeding Values
- EGVs
- Estimated genetic Values
- GEBVs
- Genomic Estimated Breeding Values
- SNPs
- Single Nucleotide polymorphisms
- GxE
- Genotype-by-environment interactions
- GxE
- Genotype-by-environment interactions
- GxG
- Gene-by-gene interactions
- GxGxE
- Gene-by-gene-by-environment interactions
- uT
- Univariate single environment one-step model
- uE
- Univariate multi environment one-step model
- MT
- Multi-trait single environment one-step model
- ME
- Multivariate single trait multi environment model
Background Genomic selection (GS) promises to accelerate genetic gain in plant breeding programs especially for long cycle crops like cassava. To practically implement GS in cassava breeding, it is useful to evaluate different GS models and to develop suitable models for an optimized breeding pipeline.
Methods We compared prediction accuracies from a single-trait (uT) and a multi-trait (MT) mixed model for single environment genetic evaluation (Scenario 1) while for multi-environment evaluation accounting for genotype-by-environment interaction (Scenario 2) we compared accuracies from a univariate (uE) and a multivariate (ME) multi-environment mixed model. We used sixteen years of data for six target cassava traits for these analyses. All models for Scenario 1 and Scenario 2 were based on the one-step approach. A 5-fold cross validation scheme with 10-repeat cycles were used to assess model prediction accuracies.
Results In Scenario 1, the MT models had higher prediction accuracies than the uT models for most traits and locations analyzed amounting to 32 percent better prediction accuracy on average. However for Scenario 2, we observed that the ME model had on average (across all locations and traits) 12 percent better predictive power than the uE model.
Conclusion We recommend the use of multivariate mixed models (MT and ME) for cassava genetic evaluation. These models may be useful for other plant species.