Evogene Ltd. (NASDAQ: EVGN) announced the completion of its generative AI foundation model version 1.0 for small molecule design, developed in collaboration with Google Cloud. The breakthrough model achieves approximately 90% precision in successful and precise novel molecule designs, representing a dramatic improvement over traditional GPT AI models that deliver approximately 29% precision.
The new model expands the existing capabilities of ChemPass AI, Evogene's tech-engine for small molecule discovery and optimization, by addressing one of the core challenges faced by both pharmaceutical and agriculture industries: identifying novel small molecules that meet multiple complex product criteria while remaining patentable.
Addressing Traditional Discovery Limitations
Traditional discovery methods typically address complex performance criteria sequentially, a process that reduces success probability and tends to steer towards well-explored or saturated areas of chemical space. This approach limits innovation potential, making it difficult to secure robust intellectual property and achieve meaningful product differentiation.
In contrast, Evogene's generative AI model enables simultaneous consideration of multiple complex product requirements while creating truly novel molecular structures. This approach facilitates the development of strong, defensible IP portfolios and paves the way for creating highly potent, synthesizable, and patentable molecules across life-science products.
Technical Foundation and Performance
The proprietary foundation model was developed in-house by Evogene's algorithm teams and built on a large dataset of approximately 38 billion molecular structures. The model was trained and deployed using Google Cloud's advanced AI infrastructure, including high-performance GPUs and scalable storage, ensuring that each compound simultaneously meets essential parameters.
"Completing our foundation model is a major milestone in our offering," stated Ofer Haviv, President and CEO of Evogene. "It unlocks new frontiers for ChemPass AI, giving us the power to generate wholly novel molecules—ones that not only perform but also create new IP space. This is key to overcoming long-standing challenges in life-science R&D: from reducing late-stage failure in pharma to developing ag-chemicals that are effective, sustainable, and proprietary."
Strategic Collaboration and Infrastructure
Boaz Maoz, Managing Director of Google Cloud Israel, commented on the collaboration: "We're pleased to collaborate with Evogene's innovation in AI-powered molecule design. Their progress with ChemPass AI highlights the strength of pairing advanced AI infrastructure with deep scientific insight. We look forward to seeing the impact of this new model in drug discovery and agriculture."
The foundation model not only powers Evogene's ChemPass AI today but will also provide a scalable base for future enhancements across the company's portfolio of subsidiaries, including Biomica Ltd. for microbiome-based therapeutics, AgPlenus Ltd. for next-generation ag-chemicals, and other life-science applications.
Future Development and Applications
Development is already underway on version 2.0 of the generative AI foundation model, with a focus on enhanced flexibility for multi-parameter optimization. The updated version will incorporate predefined, customized parameters tailored to therapeutic contexts or specific agriculture requirements, enabling ChemPass AI to better balance complex real-world constraints such as efficacy, toxicity, and stability.
This advancement will significantly improve the model's ability to generate molecules optimized for clinical, commercial, and regulatory success. Evogene welcomes continued engagement with partners across the pharmaceutical and agriculture industries interested in accessing or integrating ChemPass AI for next-generation product development.
