Protein Structure Prediction

Amino acid sequences and structural data predict 3D protein folding to accelerate drug design and novel protein creation.

Business impact

• Prediction accuracy — Higher accuracy reduces experimental validation time and increases confidence in models
• Drug discovery speed — Faster protein structure predictions shorten drug design cycles and time to market
• Designer proteins created — Enables creation of novel proteins with specific functions for biotech and therapeutics

Data requirements

• Amino acid sequences (Text) — Primary input data for predicting 3D protein folding and structure
• Known protein structures databases (Structured) — Used as training and reference data for AI model learning and validation
• Experimental structural data (X-ray, NMR, cryo-EM) (Image) — Benchmark and validate predicted protein structures against empirical measurements

AI methods and techniques

• Predictive AI — Models forecast protein folding and 3D conformations from sequence data
• Generative AI — Generates novel amino acid sequences predicted to fold into desired structures
• Agentic AI — Automates iterative design and optimization of protein structures in silico
• Symbolic AI — Incorporates physics-based rules and constraints to refine predictions

AI models and model families

AlphaFold 2, AlphaFold 3, Rosetta, ESM-2, NSP3 neural network, Variational Quantum Eigensolver

Industries

Real-world evidence

6 documented case studies on record.

Companies using this: Alchemab Therapeutics, Institute for Protein Design, Isomorphic Labs, LG AI Research, Nuclera, University of Washington.