Numerous quantitative trait loci (QTL) for various characters have recently been reported in different crop plants. However, information is limited about the molecular mechanisms behind QTL, because most of them have only been detected at a statistical level. Therefore, progeny from a cross between two soybean genotypes segregating for the presence vs. absence of the Kunitz trypsin inhibitor, a 21.5 kDa protein, have been analysed for possible effects of that protein on agronomic and seed quality characters. Protein content was reduced by, on average, 4.5 g/kg in segregants lacking the Kunitz protein, whereas oil content and other characters remained unaffected. This finding can be interpreted as a Ômodel QTLÕ for variation in seed protein content, because the molecular and genetic backgrounds of the soybean Kunitz trypsin inhibitor are well understood.

Estimating the breeding value of individuals using genome-wide marker data (genomic prediction) is currently one of the most important drivers of breeding progress in major crops. Recently, remote sensing technologies and aerial hyperspectral imaging of plant canopies combined with a variety of statistical and machine learning models allow prediction of the breeding value of individuals in the absence of genetic marker data. Hyperspectral measurements in the form of near-infrared spectroscopy have been used since the 80s to predict compositional parameters of harvest products, and some recent studies have also used this tool to predict the important parameter grain yield, suggesting that phenomic prediction can outperform genomic prediction. The genome of an individual can be considered static, however gene expression is tissue specific and differs under environmental influences, leading to a tissue and environment specific phenome. Here, we compare genomic prediction and phenomic prediction from hyperspectral measurements of wheat grains for the prediction of a variety of traits including grain yield. We show that phenomic predictions seem equivalent to genomic predictions for some traits in terms of predictive ability. However, phenomic predictions appear heavily biased towards the information present in the predictor. Future studies on this topic should investigate whether population parameters are conserved in phenomic predictions as they are in genomic predictions. Furthermore, we suggest a method to circumvent this issue, which reveals that unbiased phenomic prediction abilities are considerably lower than previously reported.