CHAI DISCOVERY

Aron Lichtman and Joshua Meier were both researchers who had worked in computational biology and saw the field transformed by AlphaFold (DeepMind's protein structure prediction model). AlphaFold solved how to predict the 3D shape of a single protein from its amino acid sequence.

The next frontier was predicting molecular interactions — how a drug molecule binds to a protein target, how two proteins interact, how a mutation changes a binding site. Chai-1, their first model, was released in September 2024 as open source and outperformed existing tools on multiple benchmarks for molecular structure prediction.

The model can predict protein-protein, protein-small molecule, and protein-nucleic acid complexes — making it directly useful for drug discovery workflows.

Chai Discovery offers a protein structure and interaction prediction platform. It generates revenue through API access for pharmaceutical companies and research institutions that need to predict how proteins and small molecules will bind — a critical step in drug discovery.

Pricing is usage-based on the number of predictions run.

Chai-1 molecular structure prediction model (open source), Chai API (commercial drug discovery access), protein complex prediction

Chai Discovery released Chai-1 as open source to drive rapid adoption in the academic and research community — a similar strategy to how Hugging Face grew by giving away models. Commercial API access and pharmaceutical partnerships are the monetization layer on top of open-source adoption.

The a16z Bio investment signaled serious backing from the most credible biotech-adjacent VC in the space.

The computational biology market is dominated by established bioinformatics companies and the academic tools that pharma companies have built internal expertise around. Chai Discovery must convince pharmaceutical companies to change their existing drug discovery pipelines to incorporate its predictions — a slow enterprise adoption cycle.

Competition from Isomorphic Labs (Google DeepMind spinout), Schrödinger, and other computational chemistry tools is intense.

Andreessen Horowitz (a16z Bio)