Google DeepMind Spin-off Isomorphic Labs Advances Drug Discovery with New AI Engine
Isomorphic Labs, a spin-off from Google's DeepMind, is making significant strides in drug discovery by developing advanced AI tools. Despite substantial investment in AI for drug development over the past decade, few AI-designed medicines have reached patients, largely due to the inherent complexity and lengthy timelines of drug testing. Isomorphic Labs, building on DeepMind's Nobel Prize-winning protein structure prediction work, has secured major partnerships with Novartis and Eli Lilly and recently raised $2.1 billion. The company has unveiled its Isomorphic Drug Design Engine (IsoDDE), a system designed to identify binding pockets on proteins and predict interactions between proteins and drug molecules.
IsoDDE goes beyond previous models like AlphaFold, which primarily focused on protein folding. While AlphaFold3 improved by modeling interactions with other biomolecules, its performance decreased when identifying novel binding sites. IsoDDE addresses this by predicting not only structure but also pocket identification and binding affinity, crucial for discovering new therapeutic mechanisms. A key validation involved IsoDDE accurately locating a previously unknown 'cryptic pocket' on the cereblon protein, a critical component in cellular protein degradation pathways. The engine also successfully predicted how ligands bind to this novel pocket, mirroring findings from a recent Nature paper.
This AI engine is expected to broaden the scope of drug discovery beyond traditional small molecules to include antibodies, molecular glues, and peptides. It aims to make more protein targets, which are currently difficult to drug due to a lack of accessible binding sites, amenable to therapeutic intervention. While acknowledging the hype surrounding AI in drug discovery, Isomorphic Labs emphasizes that a unified system like IsoDDE, with multiple predictive capabilities, is essential. The company anticipates a future where AI systems will increasingly automate hypothesis generation, testing, and analysis within the drug discovery process, moving towards agentic workflows.
The development of Isomorphic Labs' Isomorphic Drug Design Engine (IsoDDE) represents a significant step in leveraging AI for drug discovery, moving beyond structural prediction to functional interaction modeling. By focusing on identifying novel binding pockets and predicting binding affinities, IsoDDE addresses a key bottleneck in traditional drug development, potentially unlocking previously untreatable diseases. The engine's ability to generalize to new protein targets and therapeutic modalities like antibodies suggests a future where AI-driven platforms can accelerate the design and validation of diverse drug candidates. This approach aligns with the increasing need for efficient and targeted therapies in an era of complex diseases and evolving biological understanding. The challenge ahead lies in integrating these advanced predictive capabilities into robust, end-to-end drug development pipelines that can navigate regulatory hurdles and clinical validation, ensuring that AI-generated insights translate into tangible patient benefits.
AI-generated to prompt reflection — not editorial opinion, not advice, not a statement of fact. How this works.