MIT Team Gets Up to $53 Million to Develop AI-Designed Herbicides
An MIT-led research team will receive up to $53 million from ARPA-H to develop AI-designed herbicide alternatives intended to help farmers, reduce resistance and limit risks to human health and the environment.
- ARPA-H awarded an MIT-led team up to $53 million through its STREAM program.
- The PLOW project will use generative AI to design herbicide alternatives.
- Models will optimize candidates for effectiveness, cost and reduced health and ecological risks.
- The research will include high-throughput screening, greenhouse studies and microbial chemical degradation.
A new approach to weed control
The funding comes through the U.S. Advanced Research Projects Agency for Health as part of its STREAM program, or Systems for Tracking and Resilience in Efficient Agricultural-input Management. The project, called Precision Learning for On-target Weed control, will be led by Christopher Voigt, head of MIT’s Department of Biological Engineering.
More than 1 billion pounds of herbicide are used in the United States every year. These chemicals help prevent crop losses and support regenerative farming practices, but some weeds are developing resistance to widely used products. MIT also cited concerns that herbicide exposure may contribute to cancer, metabolic and endocrine disorders, and other adverse health outcomes.
“There’s a real need for new herbicides because of resistance mutations that have been arising,” Voigt says, according to MIT News. He also pointed to concerns about the possible effects of herbicides on people and the environment.
Generative AI will search for candidate molecules
The project’s generative AI team will build models that predict molecules capable of disrupting plant enzymes targeted by glyphosate and other herbicides. The candidates may be chemical compounds or small interfering RNA, also known as siRNA.
Glyphosate works by interfering with EPSP synthase, an enzyme plants need to produce several critical amino acids. According to MIT News, genetic mutations have allowed some weeds to resist herbicides after many years of use, reducing their effectiveness and threatening crop yields.
The models are expected to generate billions of possible candidates. Researchers will optimize the designs for low synthesis costs, a low risk of inducing resistance, and a low risk of harming humans or causing ecological damage. The group is led by Connor Coley and Regina Barzilay of MIT, along with Alexander Tropsha of the University of North Carolina at Chapel Hill.
Screening, greenhouse tests and chemical cleanup
A second team will screen the AI-generated candidates to identify molecules that hit their intended targets. The screening data will then be fed back into the computational models to improve them. This group will be led by Angela Koehler of MIT and Jaime Cheah of the Broad Institute of MIT and Harvard.
After the candidate pools are narrowed, a third team will conduct greenhouse studies under different agricultural conditions. The first greenhouse tests of initial RNAi candidates are expected to be complete by the end of the program’s first year.
The project will also develop living microbial systems to degrade agrochemicals before they reach food, soil or waterways. Engineered bacteria, fungi and srRNA will convert chemicals and formulation ingredients into biomass. The broader team is expected to include researchers from the Broad Institute, the University of North Carolina at Chapel Hill, the University of Tennessee at Knoxville, the University of Illinois at Urbana-Champaign and Robigo.
If successful, the work could give farmers alternatives to herbicides facing resistance problems while addressing concerns about imported phosphorus, a critical ingredient in glyphosate. The research is also designed to reduce environmental and human exposure to agricultural chemicals, according to MIT News.
