Publications des agents du Cirad


Multi-local selection of sugarcane analyzed with GGE biplots: overview of results at a glance and scope of lessons

Hoarau J.Y., Guilly S., Barau L., Thong-Chane A., Dumont T.. 2018. In : Book of abstract of the ISSCT Joint 12th Germplasm & Breeding and 9th Molecular Biology Workshops: " Improvement of sugarcane for stress environments". Okinawa : ISSCT, p. 21-21. ISSCT Germplasm and Breeding and Molecular Biology workshops. 12/9, 2018-10-22/2018-10-26, Okinawa (Japon).

Multi-environment trials (METs) represent the final stage of breeding programs prior to the commercial release of new varieties. Optimized analysis of METs impact genetic gains subsequently delivered to cane growers. The information provided by METs can be large and therefore complex to analyze and interpret when considering many environments of selection. An overview and a comprehensive interpretation ofMETs can be laborious on the sole basis of many tables of summary data and quantitative analyses of yield components. A complementary approach to interpreting many tables of figures can be obtained using "Genotype main effect plus Genotype-by-Environment" (GGE) analysis. The two-way data table of adjusted genotype means x locations is first standardized by environment. The resulting "standardized GGE matrix" of genotype main effect (G) and genotype x environment interaction (GE) is then subjected to a singular value partitioning between the genotype and environment eigenvectors. Genotypes and environments are represented on biplots defined by axes representing the most significant principal components (PCs). In order to assess effectiveness of GGE biplots to analyze sugarcane METs of Reunion Island, GGE analysis was performed on 21 sugarcane varieties tested in the MET network of eRcane that consists of seven sites of selection. These sites cover a wide range of ecologies of production representative of the main sugarcane growing areas of the industry. Varieties were assessed during two crop-cycles for tonne cane per hectare (TCH), estimable recoverable sugar (ERS), fiber content (FIB) and an economic index (EI). A biplcit represented by both PCl and PC2 : (i) adequately approximated the total GGE variation of TCH (76.52%) and ERS (71.55%) data, (ii) represented very accurately the GGE data of FIB (90.23%) and (iii) represented less efficiently the GGE data of EI (63.41 %). Such two-dimensional GGE biplots of genotypes and locations permitted to visualize at a glance: (i) congruent scatterings of genotypes on trait biplots for traits linked by significant positive (TCH and EI) or negative (TCH and ERS) correlations; (ii) a succinct summary of interrelationships among environments; (iii) best performing candidates and reliable rankings of genotypes in each environment; and (iv) the ranking of mean performance and stability of genotypes across environments. This GGE biplots graphical statistical tool permits to grasp a large scope of lessons relative to multi-local selection in the particular context of eRcane programme and provide a useful tool to rapidly assist decision-making at the time of selection.

Documents associés

Communication de congrès

Agents Cirad, auteurs de cette publication :