Protein Engineering

Leveraging geometric deep learning and structural models to predict and design functional macro-molecules.

Deep Generative Design

Traditional protein discovery is limited by the massive, unexplored search space of amino-acid sequences. Our research focuses on using diffusion models and flow-matching algorithms to design novel protein backbones.

• Geometric Representation: Representing 3D backbones as coordinates and frames (SE(3) symmetry) for equivariant network compatibility.
• De Novo Design: Generating sequence configurations that fold into desired target topologies.

Structure & Stability Estimation

Modeling contact maps and side-chain interactions to estimate structural thermodynamic stability under different chemical environments.