Evaluation of the genetic architecture of tofu traits in soybean towards genomics‐assisted breeding

Breeding soybean (Glycine max [L.] Merr.) cultivars that possess an improved suitability for tofu production is of crucial importance for tofu manufacturers. However, phenotyping of tofu‐related traits is cumbersome, costly and time‐consuming. The objective of this study was, therefore, to investigate the potential of marker‐based approaches to assist selection for tofu quality. Three recombinant inbred line (RIL) populations were evaluated for several tofu‐related traits. We found in each population at maximum one QTL per trait, explaining 10.9%–74.2% of the genetic variation, and no QTL at all for tofu weight and tofu yield. The QTL for soymilk weight, tofu hardness and tofu value was population‐specific and may correspond to QTL controlling the proportion of water‐soluble protein. The results of the QTL mapping illustrated that the tofu traits are of a highly quantitative nature. Results from the genomewide prediction were promising for seed quality traits and soaking traits, but for the tofu traits, the prediction accuracies have to be increased further based on improvements of the phenotyping procedure.