How can artificial intelligence affect agriculture, food processing and bioengineering? Scientists already have several ideas on this score. For example, a new breeding algorithm will make it possible to create ideal plant genotypes that will survive in any conditions without the need for fertilizers and special care.
For the past 5 years, a team from the Oak Ridge National Laboratory (ORNL) Department of Biological Sciences has been studying plants to understand the genetic variables and patterns that allow them to adapt to changing environmental conditions and climates. The goal of the scientists was to create an integrated roadmap for the development of AI in the field of plant breeding and bioenergy. The work was presented in the journal Trends in Biotechnology.
Scientists have developed new ways to carry out the so-called genomic selection. The genomic selection algorithm based on machine learning methods is called Explainable AI. It is an area that studies the results of AI decisions so that humans can understand how AI makes decisions.
The algorithm helps determine which genome variations need to be combined to produce plants that can best adapt to their environment. Scientists receive information about possible breeding achievements, genetic modification and the desired combinations of genes. And the data obtained already determines which bioengineering strategy should be used.
The work used the most powerful and smartest supercomputers in the world – Titan (Cray XK7) and Summit, which are located in the Oak Ridge laboratory. The latter received the Gordon Bell Prize last year for breaking the exascale barrier with a code that allows one to study combinatorial interactions between organisms and their environment.
Scientists have incorporated an AI-driven genomic selection algorithm into the Combinatorial Metrics Code (CoMet) and are now feeding it environmental information on a daily basis. This gives them the opportunity to study genome-wide associations for climatic seasons.
The researchers also built models of every square kilometer of land on the planet and encoded data on the environment and climate in them over the past 50 years, ranging from soil to the spectral composition of light.
To understand all the relationships between the different environments, they were compared to each other using a new algorithm called Duo, which was added to the CoMet codebase. This is the largest scientific calculation ever made.
According to the authors of the work, these comparisons will help determine exactly which gene mutations and alleles need to be included in the genome in order to help plants adapt to various conditions, including the most unsuitable for them.
The new algorithm will create ideal plant genotypes that thrive in adverse conditions. This, in turn, will help boost food production as well as reduce the environmental threat due to the fact that fertilizers are rarely used.