Wang, also an Iowa State Plant Sciences Institute Faculty Scholar, serves as the principal investigator for the project supported by a $2 million grant from the National Science Foundation.
Wang, along with three other Iowa State faculty members – Guiping Hu, associate professor of industrial and manufacturing systems, Sotirios Archontoulis and William Beavis, both Iowa State agronomy researchers – is creating new algorithmic models that make more accurate predictions about plant phenotypes. The research provides insight into the observable characteristics of plants, including yield, foliage color and height.
The team also includes the students of Wang’s fellow faculty members along with industry collaborators from Corteva, Syngenta and Kromite. Wang says using engineering methods to study plant genomics is an uncommon, but promising, interdisciplinary collaboration.
“Resource allocation decisions in the plant breeding process have traditionally been made through trial and error,” Wang said. “But, the efficiency of such decisions can be greatly optimized using deep Q-learning, a type of reinforcement learning algorithm.”
Wang says the challenge of the interdisciplinary collaboration is engineers’ understanding of plant genomics and plant scientists’ understanding of engineering methods. When the understanding is mutual, both engineering and plant genetics experts can engage in a productive exchange of ideas, leading to new discoveries.
With changes in global climate and environmental conditions, Wang and his team aim to optimize plant genetics to maximize resilience and expedite adaptation to changing environments.
Competition leads to superior models
Wang works to integrate his research into the undergraduate and graduate courses he teaches. Students in his industrial manufacturing and systems engineering classes get a unique introduction to the study of plant genomics and other nontraditional applications of engineering principles.
In most related research today, optimizing genomics is based on finding pairs with the highest genetic estimated breeding value (GEBV). In Wang’s class competitions, students have identified ways to further improve his methodology. A winning proposal from Wang’s students took his algorithmic models one step further by accounting for the complementarity of breeding pairs and generational differences.
“When the results were back from the student projects, I was so amazed,” Wang said. “They were really out-of-the-box designs, and they are not like any algorithm you have seen in the literature.”
The novel and innovative findings generated from Wang, his team and his students will continue to drive the future trajectory of the research, solidifying the nexus between plant genetics and engineering.