Biostimulants can improve plant growth, abiotic stress tolerance, and agricultural sustainability. Industrial side streams
contain bioactive compounds and can be feedstock for new biostimulants, yet development is slow and costly due to
greenhouse and field trials. This project uses AI and Large Language Models to unify data and build models to investigate
promising biostimulant opportunities, reduce costly in plant trials, speed innovation, and promote circularity by the use of
side streams.

Biostimulants are valuable in agriculture because they can naturally enhance plant growth, improve nutrient & water
uptake, enhance stress tolerance (e.g. to drought, salinity, heat), increase yields, and improve crop quality by stimulating
plant processes, making farming more sustainable and resilient, especially considering climate change. Various industries
(such as agri-food processing and municipal/aquaculture waste processing) yield side streams that are rich in bioactive
compounds such as polyphenols, alkaloids, and polysaccharides that can be extracted or processed into (putative) valuable biostimulants, thus promoting circular economies and converting costly waste into high-value agricultural inputs.
Research on valorising side streams through bioactive compounds currently relies on extensive laboratory work to identify useful molecules and optimize extraction of biostimulants, followed by costly and time-consuming plant growth trials, in greenhouses under controlled conditions and/or via field trials. Knowledge generated worldwide is fragmented across publications and datasets, often based on varying and diverse analysis methods, limiting accessibility and slowing
innovation. This situation hampers rapid discovery of new valorisation pathways and delays large-scale implementation of known ones. Advances in artificial intelligence (AI) offer a transformative opportunity to overcome these barriers.
This project aims to develop knowledge and AI based innovation tools for the biostimulant sector, enabling it to profitably provide for healthy soils and resilient plant production in a modern and sustainable agriculture.

This project aims to accelerate the discovery, production, and validation of effective biostimulants from side streams by
applying state-of-the-art AI. Instead of relying only on trial-and-error testing, we will consolidate existing scientific
knowledge, public datasets, and partner-provided data into a single resource. Additionally, targeted analytical
measurements can be added where most needed to strengthen the dataset.
Based on this unified knowledge base, we will develop a hybrid AI model that combines:
• data driven modelling (deriving patterns and relationships from existing datasets)
• mechanistic modelling, based on scientific knowledge of plant physiology
This integrated model will support rapid screening and prediction of new, high-performance biostimulants, helping
companies identify the most promising candidates long before investment in costly greenhouse or field trials is needed.