Chai-1

Chai-1 predicts molecular structures of DNA, RNA, proteins, and small molecules using a multi-modal foundation model.

Key Features:

• Multi-modal structure prediction: Predicts structures across DNA, RNA, proteins, and small molecules within a single framework.
• Experimental restraint integration: Optionally accepts experimental restraints derived from wet-lab data to boost prediction performance by double-digit percentage points.
• Single-sequence mode: Operates without multiple sequence alignments (MSAs) while preserving most of its performance.
• State-of-the-art drug discovery performance: Achieves state-of-the-art results across diverse molecular structure prediction tasks relevant to drug discovery.

Scientific Applications:

• Protein structure prediction: Predicts protein structures for target identification and characterization in drug discovery.
• Nucleic acid structure prediction: Predicts DNA and RNA structures to understand nucleic acid-based therapeutic mechanisms.
• Small molecule structure prediction: Predicts small molecule structures for ligand design and optimization.
• Structure-based drug discovery: Supports molecular interaction decoding across multiple biomolecular classes for therapeutic development.

Methodology:

Chai-1 employs a multi-modal foundation model architecture that integrates diverse molecular inputs and can be conditioned on experimental restraints.