Accelerate Drug Discovery: PocketGen is Now Available on Vecura

What is PocketGen?

PocketGen is an advanced generative deep-learning model designed to create full-atom sequences and sidechain structures for protein binding pockets, conditioned on a specific target ligand pose. By employing a Graph-Enhanced Transformer (GET) architecture, it simultaneously predicts amino-acid identities and precise sidechain conformations. This approach effectively resolves the primary challenge in structure-based drug discovery of engineering protein environments that are chemically and geometrically complementary to small-molecule ligands.

It helps users streamline the design of protein scaffolds tailored to specific therapeutic candidates. It is especially useful for researchers in drug discovery who need to generate protein sites that maximize binding affinity for a chosen drug molecule.

What can users do with PocketGen on Vecura?

With PocketGen on Vecura, users can:

• Design customized protein binding pockets tailored to a specific small-molecule ligand pose.
• Generate full-atom PDB structures with optimized sidechain coordinates for downstream docking analysis.
• Obtain predicted amino-acid sequences for the binding pocket to facilitate laboratory synthesis or mutagenesis studies.
• Evaluate the quality of the generated designs using integrated AutoDock Vina binding affinity scoring.

What the output means

The output provides a redesigned full-atom protein structure (PDB), the corresponding amino-acid sequence, a computed binding affinity score (via Vina), and interpretability metrics through attention weights.

This output should be used to support scientific decision making. It does not replace experimental validation.

Why this matters

The ability to computationally design protein-binding sites that perfectly match a target ligand is a transformative capability in structure-based drug discovery. Traditional methods often struggle to balance geometric precision with chemical compatibility, particularly when redesigning existing scaffolds to accommodate new ligands.

PocketGen, published in Nature Machine Intelligence (2024), represents a significant leap forward by generating sequence and structure simultaneously. This integrative approach allows for the creation of protein environments that are not just theoretically sound but are optimized for high-affinity binding, accelerating the pipeline from initial target identification to viable therapeutic candidate design.

• Developed by: Zaixi Zhang et al.
• Source: Official GitHub Repository
• Reference: Nature Machine Intelligence (2024)