Artificial Intelligence Accelerates the Development of New Soybean Cultivars - Seed World

The incorporation of artificial intelligence (AI) into soybean breeding is significantly reducing the time needed to develop new cultivars. What once required up to a decade can now be accelerated through simulations and virtual crossings that estimate how plants may perform under various environmental conditions.

Daniel Longhi, associate researcher at TMG (Tropical Improvement & Genetics), explains that AI enables the simulation of genetic combinations before field trials begin. 

“With predictive models, we can anticipate important characteristics for each region, optimizing the selection and making the launch of new varieties up to three times more agile,” he says.

Advancements in genotyping and phenotyping technologies also play a role. These tools provide detailed genetic maps of individual plants from early growth stages, according to a press release.

“We enter the DNA of the plant to accurately select the attributes that should be preserved, which increases efficiency in the field,” he explains.

TMG plans to invest R$2 billion by 2031 to expand its research and development infrastructure, with a focus on soybean, corn, and cotton breeding programs.

The use of AI in breeding supports more efficient genetic improvement and contributes to broader agricultural objectives related to productivity, crop resilience, and resource use.

Adapting Cultivars to Climate and Regional Conditions

Brazil’s diverse climate and soil conditions require region-specific genetic materials. 

“The material we plant in the South differs from those used in the Cerrado, and each region has its particularities, challenges and unique characteristics,” explains Longhi.

Field technologies, including drones for phenotyping, are being used to monitor cultivar performance under various stress conditions such as heat waves and drought.

The collected phenotypic data are integrated into genomic databases and, with the help of AI, are used to build predictive models. These models support more accurate selection and development of cultivars suited to specific environments.

“Today we can predict the performance of crossings for the next three to four years,” says Longhi. This allows strategic genetic planning, with simulations of millions of combinations to select the materials with the greatest potential for productivity and climate adaptation.