Why AI is the New Frontier for Safer Herbicides
Up to $53 million from the U.S. Advanced Research Projects Agency for Health (ARPA-H) will fund an MIT-led consortium to create generative-AI tools that design herbicide alternatives focused on human health and resistance management. The effort, part of ARPA-H’s STREAM (Systems for Tracking and Resilience in Efficient Agricultural-input Management) program, aims to reduce reliance on imported phosphorus, curb the spread of resistant weeds, and lower the exposure of farm workers and the public to potentially carcinogenic chemicals.
In short, the ARPA-H grant backs a multi-institution project that will use artificial intelligence to propose billions of new chemical or RNA-based molecules, screen them in high-throughput assays, and test the most promising candidates in greenhouse trials, all with the goal of delivering safer, more resilient herbicides for U.S. agriculture.
From Glyphosate Crisis to AI-Driven Solutions
More than 1 billion pounds of herbicide are applied annually in the United States. While these chemicals protect yields and enable regenerative practices, the over-reliance on a narrow set of modes of action—most notably glyphosate—has produced two intertwined problems. First, weeds are evolving resistance mutations that render glyphosate and similar products ineffective. Second, epidemiological studies link chronic exposure to herbicides with increased risks of cancer, endocrine disruption, and metabolic disorders. The combination threatens both food security and public health.
“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 of the Precision Learning for On-target Weed control (PLOW) project. “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 agro-chemicals that are absolutely essential for getting the high crop yields and economical prices that the United States now sees.”
Generative AI at the Core
The first of three coordinated teams will develop the computational engine. Led by Connor Coley (MIT Chemical Engineering), Regina Barzilay (MIT AI and Health), and Alexander Tropsha (UNC Eshelman School of Pharmacy), the group will build generative models capable of predicting molecules—both traditional small-molecule chemicals and small interfering RNA (siRNA)—that target the same enzymes as glyphosate, primarily EPSP synthase. The models will be trained to optimize multiple objectives simultaneously:
- High potency against target weeds
- Low likelihood of resistance development
- Minimal toxicity to humans and non-target organisms
- Cost-effective synthesis pathways
By generating “billions of possible candidates,” the AI system creates a searchable chemical space far beyond what human chemists could explore manually. The output will feed directly into the second team’s high-throughput screening pipeline.
High-Throughput Screening: From Virtual Hits to Real-World Data
Led by Angela Koehler (MIT Biological Engineering) and Jaime Cheah (Broad Institute), the screening team will synthesize or synthesize in-silico the top AI-ranked candidates and test them against the target enzymes and whole-plant assays. Crucially, the experimental results will be fed back into the AI models, creating a closed-loop learning system that continuously refines its predictions.
“Those libraries will be screened both to identify those molecules that hit their target, but also to then provide all of that screening data back to the computational approach in order to improve it,” Voigt explains.
Greenhouse Validation and Environmental Safeguards
The final team, comprising researchers from the University of Tennessee at Knoxville, University of Illinois at Urbana-Champaign, University of North Carolina at Chapel Hill, and the startup Robigo, will move the most promising compounds into controlled greenhouse studies. Led by Neal Stewart and Scott Lenaghan (UTK), Patrick Tranel (University of Illinois), Tim Willson (UNC), and Andee Wallace (Robigo), this group will assess efficacy under realistic agronomic conditions, evaluate any phytotoxic effects on crops, and begin the first tests of RNAi-based candidates by the end of year one.
Beyond weed control, the project includes a parallel effort to engineer living microbial systems—bacteria, fungi, and small-RNA constructs—that can degrade residual agro-chemicals before they enter soil, water, or food chains. This “biodegradation-by-design” strategy could further lower human exposure and environmental persistence.
Strategic Implications for U.S. Agriculture
The ARPA-H investment reflects a broader national security perspective. The United States imports a substantial share of phosphorus, a key component of glyphosate, making the nation vulnerable to supply disruptions. By discovering herbicides that either use alternative precursors or rely on RNA-based mechanisms, the project could diversify the input supply chain and reduce geopolitical risk.
Moreover, the initiative aligns with emerging regulatory trends that increasingly scrutinize pesticide safety. If successful, AI-designed herbicides could meet stricter toxicity thresholds while maintaining the agronomic performance that modern farms demand.
Potential Ripple Effects Across the Fourth Industrial Revolution
From a 4IR standpoint, the PLOW project exemplifies the convergence of several transformative technologies:
- Artificial intelligence—generative models and active learning loops.
- Synthetic biology—engineered microbes for biodegradation.
- High-throughput robotics—automated assay platforms for rapid screening.
- Data analytics—integration of chemical, biological, and agronomic datasets.
These synergies could accelerate the broader adoption of AI-driven design in other sectors, such as pharmaceuticals, materials science, and climate-resilient crops, reinforcing the Fourth Industrial Revolution’s promise of faster, more sustainable innovation cycles.
Challenges and Open Questions
While the roadmap is ambitious, several hurdles remain:
- Resistance management: Designing molecules that inherently limit resistance evolution is conceptually sound, but field validation will be needed over multiple growing seasons.
- Regulatory approval: New chemical entities and RNAi products must navigate EPA and FDA pathways, which can be lengthy and data-intensive.
- Scale-up economics: Low-cost synthesis is a design objective, yet translating laboratory yields to commercial manufacturing may reveal unforeseen expenses.
- Public perception: RNAi-based herbicides, though potentially safer, may encounter consumer resistance similar to that seen with genetically modified organisms.
Addressing these issues will require close collaboration between scientists, policymakers, and industry stakeholders—a hallmark of successful 4IR initiatives.
FAQ
What is the total amount of funding provided by ARPA-H?
The project received up to $53 million from ARPA-H.
Which institutions are part of the MIT-led consortium?
Participants include 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-Champagne, and the startup Robigo.
What is the primary scientific target for the new herbicides?
They aim to disrupt the plant enzyme EPSP synthase, the same target used by glyphosate.
How will generative AI be used in the project?
The AI models will generate billions of candidate molecules or siRNA sequences, optimizing for potency, low resistance risk, safety, and synthesis cost.
When are the first greenhouse tests expected?
Initial RNAi candidates are slated for greenhouse evaluation by the end of the first year of the program.
Will the project also address environmental contamination?
Yes, the team will develop engineered microbes that can degrade agro-chemicals before they reach soil, water, or food.
Is this effort focused on commercializing a product?
The work is currently in the research and development phase; no commercial deployment or efficacy results have been published yet.
Conclusion
The ARPA-H-backed PLOW initiative places generative AI at the heart of a new era in agro-chemical research, seeking herbicide solutions that are both effective against resistant weeds and safer for people and ecosystems. By coupling advanced computational design with high-throughput biology and greenhouse validation, the consortium aims to create a resilient, domestically secure supply chain for crop protection. While technical, regulatory, and societal challenges remain, the project illustrates how the Fourth Industrial Revolution’s toolkit—AI, synthetic biology, and data-driven experimentation—can be marshaled to confront pressing agricultural and public-health concerns.
Source: MIT News
