Statistical modeling of metabolic profiling data based on sensory attributes

To find the optimum model that can reflect the property of sensory attributes based on metabolic profiling data is one of the major challenges in plant breeding program. In this study, different regression models were compared for linking metabolic profiling to their sensory evaluation in cocoa. To achieve this goal, metabolomics data from different analytical platforms were combined to obtain a rich variety of metabolites from 76 cocoa genotypes. The sensory panel data were obtained by 11 well trained panelist score for 7 different attributes on 76 genotypes. The model used for sensory analysis is y_ijk = mu_j + beta_j * theta_i + eijk, with j= assessor, i= genotype and k=replicate in which REML and Bayesian analysis is used and compare their results. To build the prediction model, theta_i from above model used and regressed on the metabolites data