MIT-Led AI Initiative Targets Safer, More Resilient Herbicides
Researchers at the Massachusetts Institute of Technology have secured up to $53 million from the U.S. Advanced Research Projects Agency for Health (ARPA-H) to launch a multi-institution effort called Precision Learning for On-target Weed control (PLOW). The program, part of ARPA-H’s STREAM (Systems for Tracking and Resilience in Efficient Agricultural-input Management) initiative, will build generative artificial-intelligence tools that design new herbicide candidates—both small-molecule chemicals and RNA-based agents—optimised for human-health safety, low resistance risk, and affordable synthesis.
This effort aims to diversify the United States’ agrochemical supply chain, reduce reliance on imported phosphorus, and create a data-driven pipeline that can quickly iterate from billions of AI-generated structures to greenhouse-tested weed-control solutions.
Why a New Herbicide Platform Is Needed
The United States applies more than 1 billion pounds of herbicide each year. While these chemicals protect yields and enable soil-friendly regenerative practices, two converging pressures threaten their continued utility. First, weeds have evolved resistance mutations that blunt the effectiveness of legacy compounds such as glyphosate and atrazine. Second, epidemiological studies link chronic, low-level exposure to herbicides with increased risks of cancer, endocrine disruption, and metabolic disorders, affecting farm workers and the broader public through food and water.
“With herbicides like glyphosate and atrazine, there’s a real crisis because of resistance mutations that are coming,” says Christopher Voigt, head of MIT’s Department of Biological Engineering and principal investigator for PLOW. “If you combine the resistance mutations with issues around national security of the supply chain of the precursors to herbicides, we could find ourselves without these core agrochemicals that are absolutely essential for getting the high crop yields and economical prices that the United States now sees.”
How Generative AI Will Change Herbicide Discovery
The PLOW consortium—featuring MIT, the Broad Institute of MIT and Harvard, the University of North Carolina at Chapel Hill, the University of Tennessee at Knoxville, the University of Illinois at Urbana-Champaign, and the biotech startup Robigo—has divided the work into three technical streams that mirror the STREAM program’s focus areas.
AI-Driven Molecule Generation
Led by Connor Coley (MIT Chemical Engineering), Regina Barzilay (MIT AI and Health), and Alexander Tropsha (UNC Eshelman School of Pharmacy), the first team will construct generative models capable of proposing billions of candidate structures. The models will be trained to predict molecules that inhibit the same plant enzymes targeted by existing herbicides—most notably EPSP synthase, the enzyme glyphosate blocks to halt synthesis of essential amino acids.
Beyond potency, the AI will incorporate multi-objective optimisation for:
- Low synthetic cost, to keep farmer adoption feasible.
- Minimal propensity to induce resistance, by exploring chemical spaces that are less prone to mutational escape.
- Reduced human toxicity and ecological impact, guided by existing toxicology datasets.
High-Throughput Biological Screening
The second stream, overseen by Angela Koehler (MIT Biological Engineering) and Jaime Cheah (Broad Institute), will receive the AI-generated libraries and subject them to large-scale biochemical and cellular assays. Hits that demonstrate target engagement will be fed back into the computational loop, allowing the models to refine their predictions iteratively—a classic active-learning cycle.
Greenhouse Validation and Microbial Mitigation
Finally, a third team comprising Neal Stewart and Scott Lenaghan (UT Knoxville Center for Agricultural Synthetic Biology), Patrick Tranel (University of Illinois Crop Sciences), Tim Willson (UNC Pharmacy), and Andee Wallace (Robigo) will move promising candidates into controlled greenhouse trials. These studies will assess efficacy across diverse crop-weed scenarios and evaluate any off-target effects on plant health.
In parallel, the consortium will engineer living microbial systems—bacteria, fungi, and small-RNA constructs—that can degrade residual agrochemicals before they reach soil, water, or food. By converting hazardous compounds into harmless biomass, the project adds a bioremediation layer that could dramatically lower human exposure.
Potential Impacts on Food Security and Supply Chains
If successful, the AI-powered platform could address two strategic vulnerabilities highlighted by Voigt. First, by generating herbicides that do not rely on phosphorus-based precursors, the United States could lessen its dependence on imported phosphorus, a finite resource that underpins the most widely used herbicide, glyphosate. Second, a pipeline that rapidly produces resistance-proof agents would safeguard crop yields against evolving weed populations, thereby stabilising food prices and protecting the nation’s agricultural export competitiveness.
Because the project is still in its early stages—no field results have been published—the timeline for commercial deployment remains uncertain. The first greenhouse tests of RNAi-based candidates are slated for completion by the end of the first year, after which the team will evaluate whether the molecules merit further scaling.
Key Takeaways
- ARPA-H is funding up to $53 million for an MIT-led, multi-university effort to create AI-designed herbicides.
- The PLOW platform will generate billions of candidate molecules, screen them experimentally, and validate the best performers in greenhouse settings.
- Beyond chemicals, engineered microbes will be developed to degrade residual agrochemicals, adding a safety net for the environment and public health.
- The initiative targets both resistance management and reduction of human health risks associated with current herbicide use.
- Results are still forthcoming; the first greenhouse trials are expected by the end of year one.
FAQ
What is the primary goal of the PLOW project?
To develop generative AI tools that design herbicide candidates optimized for efficacy, low resistance risk, affordable synthesis, and minimal human or ecological toxicity.
How much funding does the project receive and from whom?
The effort is backed by up to $53 million from the U.S. Advanced Research Projects Agency for Health (ARPA-H) under its STREAM program.
Which institutions are participating in the research?
MIT leads the consortium, which also includes the Broad Institute of MIT and Harvard, the University of North Carolina at Chapel Hill, the University of Tennessee at Knoxville, the University of Illinois at Urbana-Champaign, and the biotech company Robigo.
What types of herbicide candidates will the AI generate?
The models will propose both small-molecule chemicals and small interfering RNA (siRNA) agents that target plant enzymes such as EPSP synthase, the same target of glyphosate.
When can we expect to see the first test results?
The first greenhouse evaluations of RNAi candidates are planned for completion by the end of the first year of the program; no field data have been released yet.
Will the project also address environmental contamination?
Yes. The team will engineer microbial systems capable of degrading agrochemicals before they enter soil, water, or food, providing a bioremediation pathway alongside the new herbicides.
How does this research relate to national security?
By reducing reliance on imported phosphorus—a key ingredient in glyphosate—the project aims to strengthen the domestic supply chain for essential agrochemicals.
Conclusion
The MIT-led PLOW initiative represents a bold convergence of generative AI, high-throughput biology, and synthetic microbiology, all aimed at reinventing the herbicide landscape. While the program is still in its discovery phase, its multi-disciplinary approach could set a new standard for how agricultural inputs are designed—prioritising human health, ecological stewardship, and supply-chain resilience. As the first greenhouse trials conclude, the industry will watch closely to see whether AI-driven chemistry can deliver the safer, more durable crop protection tools that modern agriculture desperately needs.
Source: MIT News
