Harnessing the Power of AI to Accelerate Breeding Programs

ARS scientists are employing artificial intelligence (AI) to accelerate breeding for new plant varieties that will greatly benefit farmers. Traditional plant breeding is often a slow, laborious process that includes many steps of manual assessment whose quality depends on individual observers. AI offers a promising alternative by automating these steps.

ARS scientists in St. Paul, MN, analyzed 15,000 root systems of alfalfa plants using digital images fed into an AI model to test its ability to predict root types, and they compared its accuracy and speed to human predictions. Findings showed the AI method not only can be used by non-experts because it only requires using a phone camera but can also significantly reduce the time required to identify root types from 22 to 2 weeks, while also minimizing human errors. In Beltsville, MD, ARS researchers developed a cost- and labor-efficient imaging system to monitor drought progression and recovery in 1,000 turfgrass mapping population variants. This system automatically captured 345,600 images over a month, and the AI model analysis achieved a prediction accuracy of 93%, enabling identification of key traits and associated genes affecting drought susceptibility and tolerance within the mapping population. Both AI methods delivered results in less time and at a significantly lower cost, proving to be of tremendous value for plant breeders to increase the efficiency and cost-effectiveness of their programs.

###