Terray Unveils EMMI, an Advanced AI-Driven Drug Discovery Platform

Terray Therapeutics, a biotech company specializing in AI-native and chemistry-first approaches, has announced its innovative platform, EMMI, which integrates advanced Select models to identify the most impactful molecules for drug discovery. "EMMI is a platform we have purpose built to solve some of the hardest challenges in drug discovery," stated Jacob Berlin, CEO of Terray Therapeutics.

EMMI, short for Experimentation Meets Machine Intelligence, combines Terray's proprietary ultra-miniaturized hardware for molecule testing with a fully automated lab and a sophisticated AI platform. This AI framework is underpinned by the world's most advanced chemistry foundation model, allowing for accurate Generate, Predict, and Select models. These models empower Terray's scientists to assess millions of molecules and choose the optimal candidates for experimental testing, continuously refining their models with new data.

Jacob Berlin elaborated, “Across our internal programs and for our partnership work, EMMI has allowed us to explore the dark matter of chemical space to both find and optimize novel molecules – often for targets with no known starting point. With EMMI, Terray is excited to bring a rich pipeline forward for patients.”

Terray is swiftly advancing its portfolio of immunology-focused therapeutics and collaborating with partners across various therapeutic domains. Leveraging an extensive dataset of over 13 billion target-molecule interactions, EMMI consistently identifies starting molecules for challenging targets. The platform then optimizes these molecules through a process where generative AI designs them, predictive AI evaluates their properties, and selection AI decides which to produce and test. This integration of AI with experimental methods has been shown to reduce the cost and time for iterative molecule optimization by threefold.

Terray's data advantage stems from the world's largest precise chemistry dataset, generated by its proprietary ultra-dense microarray, which produces over 1 billion new measurements quarterly. With 13 billion target-molecule measurements accumulated, Terray can uncover novel chemistry that is often inaccessible to others.

The foundation of EMMI lies in Terray’s chemistry foundation model, COATI, which is now in its third generation and trained on over a billion diverse molecules. This model, akin to a GPT for molecules, continually evolves to meet the needs of AI models by encoding molecules in a way that is understandable and manipulable by AI. “COATI3 is the world’s most advanced multi-modal chemistry foundation model and powers EMMI’s Generate, Predict, and Select models,” said Berlin.

In the drug discovery cycle, EMMI utilizes generative AI models to propose a selection of molecules tailored to meet desired properties. The generative method, recently published, has been enhanced by incorporating reinforcement learning techniques. EMMI's predictive capabilities then assess the potency and selectivity of these molecules, allowing for rapid and cost-effective predictions.

The latest advancement, EMMI's selection models, significantly enhance the efficiency of small molecule discovery. These models analyze the predictions made by the Predict models, optimizing the number and selection of molecules to create and test next. "Tested retrospectively, this means Terray can get to the key molecule 3x faster and cheaper than those using industry standard selection methods," highlighted the team.

Narbe Mardirossian, Chief Technology Officer, expressed optimism about EMMI's future capabilities, stating, "Because EMMI is used every day across our internal and partnered programs, iterative experimentation and computation continually generate insights that we incorporate directly into the platform. This systematic improvement moves us toward our vision of a closed-loop process driven by advanced reasoning models."