Classification-driven framework to predict maize hybrid field performance from metabolic profiles of young parental roots

Maize (Zea mays L.) is a staple food whose production relies on seed stocks that largely comprise hybrid varieties. Therefore, knowledge about the molecular determinants of hybrid performance (HP) in the field can be used to devise better performing hybrids to address the demands for sustainable increase in yield. Here, we propose and test a classification-driven framework that uses metabolic profiles from in vitro grown young roots of parental lines from the Dent × Flint maize heterotic pattern to predict field HP. We identify parental analytes that best predict the metabolic inheritance patterns in 328 hybrids. We then demonstrate that these analytes are also predictive of field HP (0.64 ≥ r ≥ 0.79) and discriminate hybrids of good performance (accuracy of 87.50%). Therefore, our approach provides a cost-effective solution for hybrid selection programs.