Advanced Protein-Ligand Docking with SurfDock is Now Available on Vecura

What is SurfDock?

SurfDock is a cutting-edge, surface-informed SE(3)-equivariant diffusion model designed to predict the 3D binding configurations of small molecules within protein pockets. By integrating molecular surface geometry with advanced language model embeddings for protein residues, it achieves highly accurate binding pose predictions without the need for computationally expensive molecular dynamics simulations. It serves as a powerful solution for both single-ligand docking and large-scale virtual screening workflows.

What can users do with SurfDock on Vecura?

With SurfDock on Vecura, users can:

• Perform Single-Ligand Docking: Provide a protein structure and a SMILES string to predict the most plausible binding pose of a candidate drug molecule.
• Conduct High-Throughput Virtual Screening: Rank large compound libraries by predicted binding confidence to identify potential leads for further research.
• Generate Complex Visualizations: Produce 3D PDB files of protein-ligand complexes that are ready for immediate inspection in tools like MolStar.
• Leverage Intelligent Rescoring: Utilize the integrated mixture density network (MDN) to confidence-score and rank generated poses, prioritizing the most likely binding modes.

What the output means

The output provides ranked lists of candidate binding poses, confidence scores, and 3D structural files (PDB/SDF). These results allow researchers to visualize how a small molecule sits within a protein pocket and prioritize compounds for experimental validation. While the model is highly predictive, these outputs should be treated as high-confidence hypotheses to support decision-making, not as a replacement for wet-lab confirmation.

Why this matters

The ability to rapidly predict protein-ligand interactions is a cornerstone of modern structure-based drug discovery. SurfDock significantly reduces the barrier to entry for virtual screening by providing a scalable, geometry-aware pipeline that captures critical biochemical context. By streamlining the identification of potential therapeutic candidates, it accelerates the early stages of the drug discovery lifecycle.

• Developed by: Cao et al.
• Source: Nature Methods (2024), GitHub repository
• Reference: https://doi.org/10.1038/s41592-024-02516-y