Chai Discovery

Chai Discovery has developed Chai-1, an advanced artificial intelligence model for molecular structure prediction. Launched in September 2024, Chai-1 is designed to predict the 3D structures of biomolecules including proteins, nucleic acids (DNA and RNA), small molecules, and their interactions. The model utilizes a Transformer-based neural network trained on large biological datasets and can generate predictions from sequence data alone or with additional inputs. Chai-1 has demonstrated competitive performance on several benchmarks, achieving a 77% success rate on the PoseBusters protein-ligand benchmark compared to 76% by AlphaFold3. On the CASP15 protein structure prediction task, it attained a C𝛼 LDDT score of 0.849. A key innovation of Chai-1 is its ability to make accurate predictions from single protein sequences without requiring multiple sequence alignments. The model can also incorporate experimental constraints to improve prediction accuracy. Chai Discovery has made Chai-1 freely available via a web interface for both academic and commercial use, with the model weights and inference code released as an open-source software library for non-commercial applications.

Key customers and partnerships

Chai Discovery has received support from industry partners including Dimension, Thrive Capital, OpenAI, Conviction, Neo, and Amplify Partners. The company aims to foster collaborations with research institutions and pharmaceutical companies to advance molecular biology research and drug discovery initiatives. By offering Chai-1 as a freely accessible tool, Chai Discovery seeks to promote widespread adoption and shared knowledge benefiting all stakeholders in the ecosystem.