The Application Of Biplot In Cultivar Evaluation And Genotype-by-environment Interaction Analysis

In crop genetics and breeding research, biplot is always used in genotype com-parison, environments classifcation and genotype-by-environment interaction (GE)analysis. It plays a major role in analyzing multi-environment trials data. But dur-ing the development of the biplot, some unreasonable results caused by various newmethods which is lack of theoretical basis or is used incorrectly make some breedersdoubt the scientifc nature of the biplot. This paper bases on the theory of singularvalue decomposition (SVD), principle component analysis (PCA) and biplot, exactlydiscusses the building and analytical methods of GGE and AMMI. There are fourmajor parts in this paper.(1) Theoretical part. In this part, the paper proves the consistency of SVD andPCA introduces the application of PCA in biplot and discusses degree of ftting.(2)GGE model part. In this part, the paper uses a real yield trial data to demonstratedetails of GGE model’s and GGE biplot’s establishment, and how can we use themin special adaption, general adaption, genotype similarity analysis and environmentevaluation. It points out that the existing method of cultivars stability analysis usedby GGE is lack of theoretical basis, and it gives correct procedure of genotype stabil-ity and genotype similarity analysis. It also clears the computation of environmentdiscrimination parameter Djand cultivars stability parameter Di, and generalizesall above procedures into multidimensional situation.(3) AMMI model part. In thispart, the paper uses the same yield trial data to demonstrate specifc details of AMMImodel’s and AMMI2biplot’s establishment, and how can we use them in GE analy-sis, special adaption, general adaption and genotype similarity analysis. It reasonablyevaluates AMMI1biplot and improves it into “general adaption analytic diagram”.The paper also explaines why AMMI2biplot is not suitable for special adaption anal-ysis and the consistency of genotype stability parameter Dicalculated from AMMIand GGE. Finally, it generalizes all above analysis procedures into multidimensionalsituation.(4) Simulation part. In this part, the paper obtains a conclusion throughsimulation that the accuracy of genotype estimation in AMMI model is higher thanit in GGE model.AMMI and GGE have overlapping function. But There are no contradiction between them. The few diferences are accuracy, possibility of diagrammatic analysisand way of information extraction. At the same time, AMMI and GGE have theirunique felds. Environment evaluation can only be carried out in GGE model, andGE analysis only in AMMI model. So they cannot replace each other. In order toget more accurate results, this paper suggests choosing optimal models and biplotsunder diferent purpose. This paper afrms the scientifc nature of biplot and hasguiding signifcance in multi-environment trials data analysis.