The quest to develop hardier, higher-yielding crops has always been a driving force in plant breeding. Plant breeders, for generations, have relied on a combination of selection and environmental adaptation to achieve these goals. But in the face of a changing climate and evolving environmental conditions, this task becomes more complex.
“Understanding how climate data interacts with genetics to influence plant traits has become a crucial aspect of modern plant breeding,” says Dr. Rupashree Dass, research manager at Computomics. That means new tools are needed to help breeders integrate climate data into their breeding programs.
“Plants, like humans, develop variations based on the climate they exist within,” says Dass. The interplay between genetics and environment is at the heart of what makes each living organism unique.
“For plants, changes in phenotype occur due to genetic effects, environmental effects, field management effects or a combination of all of these. Different locations with varying climates, including factors like temperature, humidity, sunlight, and rainfall, play a significant role in shaping the phenotype or phenotypic traits of a plant.”
This interaction between genetics, environment and field management is often referred to as G×E×M, signifying the interplay of genetics (G), environment (E) and management (M). It’s crucial to understand how a plant will grow and perform in a specific environment, considering variables like extreme temperatures, varying levels of rainfall, and soil types.
Machine learning comes into play by enabling plant breeders to make more informed decisions based on vast amounts of data. It can also be applied to model future climate scenarios for specific locations. This information can then be used to identify crop types that will be well-suited to those new conditions, or to identify where favorable conditions for specific crops may emerge in the future.
Dass highlights the use of machine learning in predicting plant performance in unknown locations with unique climates.
“Let’s say we want our breeding program to move to Texas. We would then like to understand, which genetics would perform the best in the conditions there. Machine learning can help us simulate the environmental conditions and predict the performance of a crop in new or unknown locations, placing the right product in the right environment. ”
Furthermore, machine learning aids in identifying which specific genes are critical for particular traits and how they interact with the environment to yield those traits.
“Understanding the genetic basis of these traits can accelerate the breeding process significantly,” she adds.
Computomics’ AI-powered solution ⨉SeedScore® offers plant breeders an innovative approach. It’s an artificial intelligence platform tailored to the world of plant breeding.
“It is a tool to find the best-performing genetics for a specific location or climate. Additionally one can use it to predict the performance of novel genetics in known locations.” Dass says.
⨉SeedScore® can predict the performance of plant genetics in various environments, making it an invaluable asset for breeders looking to expand their programs to new locations or adapt to changing climates. It helps identify which genetics will thrive in a particular region, ensuring that breeders can make informed decisions about where to focus their resources.
Dass shares a compelling example of a publicly funded project (HEB-KI) involving Computomics and Martin Luther University Halle-Wittenberg in Germany. The project aims to predict the flowering time of a global barley population in various locations from four continents.
“Using both genetic and environmental data, the project demonstrates the power of machine learning in predicting plant performance in previously unexplored regions. This project showcases the potential impact of integrating climate data and genetics in modern plant breeding,” she says.
While the integration of machine learning and climate data offers remarkable opportunities, there are also several challenges. One such challenge is convincing traditional breeders of the advantages of machine learning.
“It requires a shift in mindset and an understanding that machine learning is a tool to assist, not to replace breeders”, Dass says.
Another significant challenge is data quality.
“Collecting accurate and comprehensive historical data, including genetic, environmental and management information, is crucial for the success of machine learning models. Breeders may need to tweak their data collection practices in order to leverage the full potential of this technology.”
Machine learning and climate data integration can benefit breeders whose programs are dependent on knowing how certain genetics will behave in specific climate conditions. The more diverse the available data, the more powerful the machine learning models become.
Although the technology already delivers remarkable results, it is constantly being improved. Currently the research team at Computomics is putting a lot of effort into improving the AI models to determine resilient genetics that can perform well under most environmental conditions. Another focus lies in predicting the performance of untested genotypes and in optimizing genetic diversity.
Dass emphasizes that every breeding program is unique, and there is no one-size-fits-all solution. “Customization is key to addressing the specific challenges and goals of each client. We always work jointly together with our clients as understanding the needs of their breeding programs is important.”
She believes that as the technology evolves, the potential for revolutionizing plant breeding is enormous, potentially reducing the time it takes to bring new crop varieties to market and adhering to sustainable agricultural practices.
More information on ⨉SeedScore®.
Dr. Rupashree Dass is Research Manager at Computomics.
Contact Rupa at firstname.lastname@example.org
Originally posted on 23 October, 2023 on SeedWorld