Evogene, an AI-driven biotech company based in Israel, has launched a powerful generative AI model for designing novel small molecules in pharmaceuticals and agriculture. Developed in collaboration with Google Cloud, this marks one of the most sophisticated applications of foundation models in biotech to date. The platform is designed to generate compounds with specific properties, reduce the need for wet-lab screening, and fast-track the path from idea to lead molecule.
The model is trained on millions of compounds with annotated bioactivity data and can now generate drug-like molecules optimized for a range of parameters including solubility, toxicity, synthetic accessibility, and target specificity. According to internal benchmarks, the model achieves over 90 percent predictive success in silico before any experimental validation.
Features Designed for Bio-Developers
Evogene’s new platform is built to empower developers working on early-stage drug discovery. It offers an intuitive REST API and a Python SDK, allowing seamless integration into existing cheminformatics pipelines.
from evogene_bio import MoleculeDesigner
designer = MoleculeDesigner(api_key="YOUR_API_KEY")
molecule = designer.generate(
target="EGFR",
properties={
"solubility": ">100µM",
"toxicity": "low",
"synthesizable": True
}
)
print(molecule.smiles)
Users can also upload custom datasets to fine-tune the model to proprietary targets or screening results, and filter outputs through predictive ADMET profiling.
Integration and Scalability
The backend leverages Google Cloud’s Vertex AI platform, enabling massive scale-out capabilities for pharmaceutical companies or academic groups with high-throughput design needs. Molecules can be ranked by predicted efficacy, clustered by scaffold similarity, and exported in SMILES or SDF format for further simulation.
Evogene's team notes that the model is not limited to drug discovery but is also being applied to agrochemical development, offering greener alternatives to existing pesticides and herbicides by identifying novel, biodegradable compounds.
Why This Matters
This release exemplifies the shift toward multi-purpose generative AI in life sciences. By bringing a single foundation model to multiple domains, Evogene is building a platform that could reduce the cost and time of discovery efforts by orders of magnitude. While foundation models like AlphaFold revolutionized structure prediction, Evogene’s tool aims to tackle the front-end of drug and chemical design with the same level of disruption.
Evogene has made access to the platform available for select research collaborators and is expected to open up broader developer access by the end of 2025.
Sources
https://www.businesswire.com/news/home/20250605005633/en/Evogene-Announces-AI-Drug-Discovery-Model
https://www.techlifesci.com/p/weekly-techbio-highlights-44-new
https://www.evogene.com/newsroom
Top comments (0)