Farmers face an ever-changing combination of diseases, pests, and climate change as they work to feed the world. This means the challenge of providing them with advanced ways to protect their crops sustainably increases year-on-year. 

In 2019, Syngenta scientists envisioned using artificial intelligence (AI) to design the next generation of crop protection agents. They aimed to develop new, highly effective molecules at unprecedented speed, while meeting stringent safety and sustainability criteria.

The collaboration between IBM and Syngenta has been a significant step forward, enhancing the stage of the invention process that involves synthesizing new active ingredients.
 

Introducing Natural Language Processing to chemistry

During three years of collaboration, IBM and Syngenta have successfully combined IBM's advanced Natural Language Processing (NLP) tools with the vision of Syngenta's research chemistry scientists.

This initiative began when IBM and Syngenta scientists demonstrated NLP's capability to learn Syngenta's reaction "dialect" and make accurate reactivity predictions in its largely unpublished chemical reaction space.

AI at the center of synthesis design

The partnership began after Marco Stenta, a Syngenta Science and Technology Fellow, contacted IBM Research.

"IBM disclosed their scientific idea as an open-source Python package. This allowed me to test the approach before reaching out to the IBM team. After some years of intensive work, AI is now at the core of our synthesis design – all new processes go through it,” says Marco.  

Efficiencies gained include fewer experiments, increased ‘right-first-time’ rates, and lower costs.

"This efficiency level means that our chemists can delegate repetitive tasks to machine learning while focusing on more strategic, high-value work," adds Marco.

Teodoro Laino, Distinguished Research Scientist and Manager at IBM Research, reflects: “Everyone now knows about language models but, back in 2019, this was a very far-sighted approach from Syngenta.”

A six-month preliminary collaboration later, and the partnership was underway.

“It’s been an amazing collaboration and an awesome experience,” says Teodoro.

Openness was “in the DNA” of the collaboration, and key to its success, adds Marco. “We established an open collaboration with robust data protection practices: we protected IBM’s software and our chemical data. The collaboration was so smooth, and we enthusiastically embraced it.”

The incorporation of AI into Syngenta's new chemistry design processes will benefit more than just its chemists, with accelerated design of more sustainable synthesis procedures helping Syngenta meet the needs of growers.

Monika Trehan, Syngenta's Digital Collaborations Lead, observes: "With this approach, we can bring superior compounds to the market faster, benefiting farmers directly. Society also profits from our ability to minimize waste and find the most sustainable compounds."
 

A model collaboration

The IBM experience has set the standard for future Syngenta collaborations.

Marco remarks: "Our approach to digital collaborations has evolved massively after this experience with IBM. We feel we can be more open, sharing and amplifying ideas in pre-competitive spaces."

Teodoro concurs: "The partnership with Syngenta enabled us to apply our modelling approach to proprietary data, which would have been otherwise inaccessible. It not only highlighted real-world problems but also helped us improve our platform. This work showcases how these reactivity models can be tailored to industrial partners' needs, delivering immense value through increased predictive power."

Syngenta's IT team is now fine-tuning the platform integration details, while data scientists are planning the next phase, including the modeling of bio-catalyzed reactions and metabolic degradation pathways.