Advanced Biomolecular Structure Prediction with OpenFold3 on Vecura
This update enables structural biologists and drug discovery researchers to predict complex biomolecular structures through a guided workflow inside Vecura, without setting up complex technical infrastructure.
What is OpenFold3?
OpenFold3 is an open-source, PyTorch-based reimplementation of the groundbreaking AlphaFold3 architecture, now optimized for seamless deployment as an NVIDIA NIM microservice. It utilizes a sophisticated diffusion-based model to iteratively denoise atomic positions, enabling the accurate prediction of all-atom 3D structures for diverse biomolecular complexes. It is especially useful for researchers needing to model complex interactions between proteins, DNA, RNA, and small-molecule ligands in a single inference pass.
What can users do with OpenFold3 on Vecura?
With OpenFold3 on Vecura, users can:
- Predict the all-atom 3D structure of any biomolecular complex including proteins, DNA, RNA, and ligands.
- Co-fold protein–ligand systems without requiring separate docking procedures.
- Generate multiple diverse structure candidates to explore conformational uncertainty using diffusion sampling.
- Evaluate prediction reliability through integrated confidence scores like pTM, ipTM, pLDDT, and pDE.
What the output means
The output provides an all-atom 3D structure file (in mmCIF or PDB format) alongside a comprehensive suite of confidence metrics. These scores—including global fold quality (pTM), interface accuracy (ipTM), and per-residue confidence (pLDDT)—allow researchers to quantitatively rank diffusion samples and assess the structural reliability of the generated model.
This output should be used to support scientific decision making. It does not replace experimental validation.
Why this matters
The ability to accurately predict the interaction between various biological molecules is a fundamental challenge in drug discovery and structural biology. By moving beyond proteins to include DNA, RNA, and ligands, OpenFold3 offers a more holistic view of the cellular machinery, potentially accelerating the design of novel therapeutics and our understanding of molecular mechanisms.
Accessing this through a managed infrastructure like Vecura lowers the barrier to entry, allowing scientists to leverage high-performance GPU inference without the complexities of managing specialized containerized environments.
- Developed by: AQ Laboratory (OpenFold) based on the architecture by Google DeepMind.
- Source: Official GitHub repo (aqlaboratory/openfold-3)
- Reference: AlphaFold3 paper (Nature 2024)
Try OpenFold3 on Vecura.
Open the model workspace and start evaluating it with your own inputs.