Deep machine-learning and dynamic profiling can maximize bread quality of Einkorn and Emmer hulled wheat species

Authors: Jaradat, Abdullah; Dykes, Linda

Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 11/5/2019
Publication Date: N/A
Citation: N/A

Interpretive Summary:

Technical Abstract: Growing consumer demand for organically-produced and healthy wheat products from hulled wheat species [e.g., einkorn, Triticum monococcum L. subsp. monococcum; and emmer, Triticum turgidum subsp dicoccon (Schrank) Thell.] revitalized the interest in searching for, selecting germplasm and breeding new cultivars with high-quality and balanced kernel, flour and bread composition. Kernel, flour, and bread samples of spring-type germplasm representing a wide range of regional sources (17 countries in The Fertile Crescent, East Africa, East Europe, and West Europe) and improvement status (landraces, old cultivars and improved germplasm) of both species were analyzed for their ionome (macro- and micro-nutrients, especially iron and zinc contents); protein composition; rheological qualitative and quantitative dough properties; and bread physico-chemical composition and color space coordinates. A relational database was based on 3-year research project at the Swan Lake Research Station, Morris, MN; and was mined for multi-trait variation, association and functional relationships. A machine deep-learning algorithm was used to identify common or species-specific traits with significant roles in defining and predicting bread quality as a "latent variable." Inter- and intraspecific variation in quality-related traits due to source of origin (i.e., regions and countries within regions) may have been caused by the interplay between local preferences, management and environmental conditions. Larger overall variation in the ecologically wide-spread emmer was explained by differences among regions than in the more geographically-restricted einkorn. A dynamic profiling procedure was employed to select genotypes from both wheat species in order to maximize bread "quality" as compared to that of a spring bread wheat reference cultivar. Maximized quality index of the final product (i.e., bread) was based on quantitative adjustments of traits’ complementarity (e.g., large Fe, Zn, carotenoids and b* of einkorn from Turkey in the Fertile Crescent; and large kernel weight and loaf volume of emmer from Ethiopia in East Africa) between both wheat species. A module demonstrating deep machine learning and dynamic profiling will be presented.