2nd Nature and Kanematsu launch AI-driven discovery program to turn food waste into functional ingredients

In food manufacturing, waste is rarely empty. It is full of proteins, fibers, enzymes, and small molecules that never make it beyond the margins of production. The challenge has never been whether value exists in these side streams, but whether it can be found, understood, and turned into something usable at scale.

In response, 2nd Nature has partnered with Kanematsu Corporation to tackle that problem directly, applying artificial intelligence to analyze agricultural and food processing byproducts and identify compounds that can be developed into functional ingredients.

• 2nd Nature has partnered with Kanematsu Corporation to use AI to analyze agricultural side streams and identify functional ingredients including proteins, fibers, and bioactives.
• The collaboration has begun with a discovery phase focused on mapping molecular potential and generating a shortlist of commercially viable compounds for further development.
• The partnership has aimed to convert food processing byproducts into scalable ingredient supply chains using 2nd Nature’s end-to-end Intelligent Upcycling Platform.

The collaboration begins with the first stage of 2nd Nature’s platform, described as ‘Discovery’, where its AgWaste Portal analyzes the molecular composition of side streams and predicts how individual compounds will behave in real-world applications.

“Our AgWaste Portal is designed to look for ‘hidden’ high-value compounds, such as novel proteins, functional fibers, natural sweeteners, and bioactives,” said Dr Effendi Leonard, Co-founder & CEO of 2nd Nature, speaking with Protein Production Technology International following the announcement yesterday. “Our primary focus is on predicting functional properties. We want to know how a specific molecule will perform, taste, and feel in a final product.”

“Our primary focus is on predicting functional properties. We want to know how a specific molecule will perform, taste, and feel in a final product”

That focus marks a shift from traditional approaches to ingredient discovery, which often prioritize identification before application. Here, the emphasis is inverted: functionality comes first, and only then does a compound become relevant.

“By prioritizing functionality from the outset, we ensure that the compounds we discover have real-world commercial viability, rather than just being scientifically interesting,” Leonard continued. “Concentration is also an important aspect when considering unit economics, i.e. we prioritize ingredients that are present in the raw materials at sufficiently high amounts.”

For Kanematsu, a trading company with operations spanning food, materials, and industrial sectors, the collaboration offers a way to reassess existing supply chains through a different lens. The company has sought to understand how underutilized byproducts might contribute to more sustainable resource use.

“At Kanematsu, our corporate principle is to contribute to a sustainable society through innovation,” added Kazuhiro Matsuura, General Manager of the Growth Strategy Office at Kanematsu Corporation. “Through this collaboration, we look forward to learning from 2nd Nature’s AI-driven approach and gaining new perspectives on how agricultural side streams may contribute to more sustainable resource utilization.”

Across the global food system, large volumes of nutrient-rich material are routinely treated as low-value outputs or disposed of entirely. 2nd Nature’s approach reframes these streams as potential sources of high-performance ingredients.

“Food processing side streams are treasure troves full of nutrients and high-value ingredients. For too long, they are viewed as waste,” Leonard said.

The immediate objective of the partnership is to generate a shortlist of target compounds from Kanematsu’s side streams. That shortlist is produced by screening millions of molecules and ranking them based on predicted functionality and commercial potential.

“The goal of the discovery phase with Kanematsu is to map the molecular potential of their specific side streams and emerge with a shortlist of target compounds,” Leonard said. “We are using our AI to screen millions of molecules to identify those with the highest potential for upcycling into premium functional ingredients.”

“However, because our platform is built for end-to-end commercialization, this shortlist serves as the foundation for a defined product pathway,” he said. “The insights generated here will directly inform the subsequent development and deployment phases, ultimately turning these side streams into resilient supply chains.”

Turning that shortlist into something that can be produced reliably is where the technical challenge begins. Agricultural side streams are inherently variable, with composition influenced by factors such as seasonality, geography, and upstream processing conditions.

“Variability is a known challenge in certain agricultural side streams, but our vertically integrated approach is designed to handle it,” Leonard said. “At the production stage, our parallel downstream processing hardware tests multiple processes simultaneously.”

“This allows us to identify the most robust and optimal methods that maximize yield and quality, even when the input material fluctuates. We engineer resilience into the process from day one.”

“The filter from ‘interesting’ to ‘commercially relevant’ is built directly into our Intelligent Upcycling Platform”

Rather than relying on step-by-step optimization, which can slow development and increase cost, 2nd Nature runs multiple process configurations in parallel. The aim is to identify scalable pathways more quickly and reduce the risk of failure later in development.

“The real technical challenge often lies in developing a scalable, cost-effective process from the lab to commercial production,” Leonard said. “Traditional practices toward upscaling rely on sequential improvements, which is expensive and time-consuming.”

“We circumvent this by using our parallel downstream processing hardware to test multiple methods simultaneously. This ‘more shots on goals’ approach ensures that we not only rapidly identify a process which produces the target compound efficiently but also stabilize it so that its functional properties, whether that is taste, texture, or nutritional benefit, hold up perfectly in the final consumer product.”

Alongside technical feasibility, commercial filters are applied early in the process. Compounds are evaluated not just on performance, but on whether they can meet market demand, achieve cost competitiveness, and navigate regulatory requirements.

“The filter from ‘interesting’ to ‘commercially relevant’ is built directly into our Intelligent Upcycling Platform,” Leonard said. “Once the AI identifies top candidates based on predicted functionality, we evaluate them against key commercial metrics: functional performance, market demand, unit economics, and regulatory path.”

“We are looking for solutions to the food industry’s most pressing challenges. For instance, if a compound can serve as a superior non-calorie natural sweetener relative to other in-market sweeteners, and if it can be produced cost-effectively at scale, it moves to the top of our priority list for experimental validation.”

The platform is not limited to a single category of ingredients. Instead, it is designed to operate across a broad range of compound types, depending on the characteristics of the side stream being analyzed.

“Our AgWaste Portal is designed to be agnostic in terms of its ability to identify different classes of compounds,” Leonard said. “Specifically, it’s capable of uncovering high-value small molecules, peptides, enzymes, proteins, and functional fibers.”

“In general, the specific compounds vary depending on the side stream, whether it is grains, legumes, fruits, vegetables, or fermented dairy by-products,” he said. “However, the common thread is that these compounds are geared toward high-demand applications in the Food and Beverage, Home and Personal Care, and Wellness markets.”

Early work has already surfaced examples of commercially relevant compounds across these categories.

“For example, we have already seen great success in identifying zero-calorie sweeteners, non-sodium umami enhancers, and detergent enzymes that belong to small molecule, peptide, protein, and enzyme classes,” Leonard said.

“The real technical challenge often lies in developing a scalable, cost-effective process from the lab to commercial production”

As the collaboration progresses, the focus will shift from identifying compounds to integrating them into production systems and product formulations. A key consideration is how easily these ingredients can be adopted by manufacturers without requiring significant changes to existing infrastructure.

“Our ultimate goal is to provide manufacturers with a de-risked, resilient supply of upcycled ingredients sourced directly from their own side streams,” Leonard said. “To achieve this, the processes we develop are designed to be as seamless as possible to integrate into existing manufacturing systems.”

“We also aim to create ‘drop-in’ solutions that can slot into existing CPG formulations without requiring massive overhauls of their manufacturing infrastructure.”

The broader ambition of the platform is to connect discovery directly to deployment, shortening the distance between identifying a compound and producing it at scale. By linking molecular analysis with process development and manufacturing integration, 2nd Nature is attempting to build ingredient supply chains that originate within existing production systems.

“Our AI looks for valuable ingredients within waste streams,” Leonard said. “While our work with Kanematsu begins with mapping molecular potential, our platform is built to eventually help partners turn their biggest cost centers into resilient supplies of safer and healthier ingredients, as well as new revenue streams.”

For now, the outcome of the discovery phase will determine which compounds move forward. But the underlying premise is clear: the next generation of ingredients may not come from new crops or inputs, but from materials that have been present all along, waiting to be understood at a molecular level and engineered into something useful.