Multi-trait genomic prediction to improve tomato aroma contents

International audience

anglais. The increasing consumer demand for tastier tomatoes led to the development of several breeding projects for quality. The first ones are based on increasing the sugar content and improving the acid balance (malic and citric acids) as both traits have been widely described as the main contributors to tomato flavor. However, sugar content and fruit weight are primarily regulated by linked genetic regions with opposite effect, so increasing sugar content reduces fruit weight. In this context, other breeding levers have been considered: the most promising is the diversification of tomato aromas. Tomato aromas are combinations of large number of volatile compounds among which 30 have a major effect on taste. Each of these is controlled by a large number of QTLs. Genomic prediction is a statistical method to estimate the breeding value of genotypes based on their genotyping information. It is well suited for the improvement of low effect polygenic traits. Thus, we explored the efficiency of this method for aroma content prediction on GWAS data previously published by Bineau et al. (2021): 44 volatiles assessed by GS-MS in a collection of 121 cherry tomato lines. The ability and accuracy of prediction models adapted to multiple traits were tested by cross-validation in comparison with single trait models. The second challenge was to predict the breeding values of hybrids (derived from the line collection). Third, we explored some index based on combined volatiles content representing aroma to predict global tomato flavour. I will present the latest results on these three axes.