Dealing with Statistical Interactions in Plant Breeding within the LABKEY Software

Plant breeders have to deal with at least two fields of quantitative interactions during their decision making process: most importantly, genotype by environment (G*E) interactions, and for hybrid crops, specific combining abilities (SCA) which represent the interactions between the male and female parents of each hybrid. Moreover, these two fields are often overlapping. Based on well-known concepts which, however, are not always implemented in a ready-to-use toolkit, LABKEY is offering solutions aiming at providing various graphical views to breeders in order to help them to explore the space of these interactions. Most of these views are intrinsically based on factorial regression analysis, using either (i) an external environmental covariate, when available, or (ii) covariates derived from the data, e.g. factorial variables obtained from a principal components analysis (PCA),  the latter approach making the interpretations more difficult.  Regarding G*E interactions, graphical representations of the various components of interest – yield potential, responses to an environmental limiting factor – are proposed. For hybrid crops, the genetic backgrounds of the lines in one of the two parental sets of a factorial cross design constitute a sort of environment for the other parental set. Hence, this toolkit proposes to handle such data with the same descriptive, statistical approaches, thus allowing to reveal promising or unpromising partners. A workflow accounting for the two fields of interactions (G*E, SCA) is exemplified.