The use of historical datasets to develop multi-trait selection models in processing tomato

Abstract

AbstractMulti-trait indices (MTI) weigh traits based on their importance to facilitate selection in plant and animal improvement. In animal breeding, economic values are used to develop MTIs. For vegetables, economic data valuing traits are rarely available. We posit that varieties with traits valued by growers and processors achieve higher market share and longer life span. Our objective was to develop MTIs predicting success of tomato varieties. Historical data for the California processing tomato industry from 1992 to 2013 provided measurements for yield, soluble solids (Brix), color, pH, market share, and life span for 258 varieties. We used random models to estimate best linear unbiased predictors (BLUPs) for phenotypic traits of each variety, and evaluated trends over time. Yield has been increasing from 2006, while Brix stayed constant. Because yield and Brix are negatively correlated, this trend suggests that Brix influenced selection. The average number of resistances reported in varieties ranking in the top ten increased from 2 to 4.5 between 1992 and 2013. MTIs predicting success from phenotypic traits were developed with general linear models and tested using leave-one-out cross validation. MTIs weighing yield, Brix, pH and color were significantly correlated to success metrics and selected a significantly higher proportion of successful varieties relative to random sampling. The index multiplying yield and brix, suggested in the literature, was not significantly correlated with variety success. The MTIs suggested that fruit quality had less of an influence on variety success than yield. The MTIs developed could help improve gain under selection for quality traits in addition to yield.