We consumers are really good at identifying single differences between foods, especially if they are labeled with bright colors, but we are not very good at evaluating whether they are good for our health. This problem is made worse by the massive noise from the nutritional nonsense industry. For example, the relationship between oxidative processes, DNA damage, aging and antioxidant foods is more complicated than often advertised (Sci. Am. 308, 62–67, 2013). Plant breeders are really good at producing more calories from grain and more biomass, but they are only recently addressing the competing consideration to provide better nutrition. Plant geneticists can readily engineer high levels of single essential nutrients into staple crops such as rice. However, in the case of beta-carotene–enriched 'Golden Rice' (http://www.goldenrice.org/), implementation has stalled on the mismatch between the global demand for nutrition and the lack of clinical studies to support biofortification of food crops with single nutrients (Nat. Biotechnol. 25, 623, 2007). This situation was aptly summarized by Joseph Hirschberg at our recent conference on crop yield (http://www.cropyield2016.com/): “We in plant biology, plant physiology and plant biochemistry know more about how the compounds are produced and how to manipulate them than the nutritionists know about what we need.”

The domestication history of carrots (pages 589 and 657) includes selection for a rainbow of spontaneous mutants that enrich this root crop with purple anthocyanins, red lycopene, yellow lutein and orange beta-carotene, all of which act upon the human body and influence the nutrition we get from eating carrots. But these traditionally bred variants are a tiny fraction of the diversity we might have, now that we can make nearly isogenic crops with contrasting metabolic profiles via transgenics (Nat. Commun. 6, 8635, 2015), gene editing and marker-assisted breeding. To see whether these crops grow better and produce more yield, randomized-block field trials can be used, also testing nutrient and water use and the resistance of the crop variants to stressors, including diseases and competitor plants. Then, clinical intervention trials can establish whether there is any truth that 'eating the rainbow' is good for health or whether by selecting foods of different colors we just avoid overreliance on calorie-rich staples and spread our risk of consuming antinutrients.

Further to this point, genomic plant breeding is now enabling an understanding of how antinutrient compounds (including saponins, alkaloids, phytate, tannins, oxalate and lectins) that we might want to breed out of crop plants affect the performance of these plants in competition with weeds and fungi and predation by insects and nematodes. An excellent example of using genomics to understand the generation of crop antinutrients—that also offers a direct genetic strategy to remove them—addresses the alkaloid synthesis gene clusters in the Solanaceae (Science 341, 175–179, 2013).

Following our crop yield conference, a number of action points emerged, the most ambitious of which was better understanding and prediction of the ways in which crop yield and nutritional improvements result in compromise with other agronomic traits (tradeoffs) and identification of those cases where the genetic architecture can be altered to deliver better crops without compromising growth or stressor resistance. We also have a wish list of improvements to the sciences supporting plant breeding. First, we urge the development of standards for in vivo demonstration of the nutritional and nutraceutical properties of crop plants. In human and animal trials, it would seem to be mechanistically important to distinguish between direct metabolic effects and those mediated via the microbiome. For more rapid progress in the field, it would be good to erase the distinction between biochemical properties gained by transgenesis and by domestication and selection of mutations. The end goal of this research is to ensure that yield optimization by plant breeding includes consideration of nutritional optimization, not merely gain in calories or biomass.