ImmuneBuilder is now available on Vecura: Rapid 3D structure prediction for antibodies, nanobodies, and TCRs

What is ImmuneBuilder?

ImmuneBuilder is a suite of three specialized deep-learning predictors developed to predict the 3D structures of immune receptor proteins from their amino-acid sequences alone. The suite includes ABodyBuilder2 for paired heavy/light-chain antibodies, NanoBodyBuilder2 for single-domain nanobodies, and TCRBuilder2+ for T-cell receptors. It helps users quickly generate high-accuracy, all-atom structural predictions of the variable domains of these proteins. It is especially useful for high-throughput therapeutic design pipelines where structural modeling of hundreds or thousands of candidate sequences is required.

What can users do with ImmuneBuilder on Vecura?

With ImmuneBuilder on Vecura, users can:

• Predict paired antibody structures: Generate 3D all-atom structural models of paired heavy- and light-chain variable domains (Fv) using ABodyBuilder2.
• Model single-domain nanobodies: Predict the structures of camelid or synthetic VHH nanobodies using NanoBodyBuilder2, requiring only the heavy-chain sequence.
• Determine T-cell receptor structures: Leverage TCRBuilder2+ to predict paired alpha/beta TCR structures for epitope mapping or profiling immune repertoires.
• Run energy-minimized refinements: Perform automated OpenMM/pdbfixer energy minimization on predicted models to ensure biophysically realistic structures.

What the output means

The output provides a predicted 3D all-atom structure in PDB format, an overall mean error estimate, and per-residue RMSD error estimates in Angstroms. This structure file can be easily visualized in macromolecular modeling suites, while the error estimates help users assess model reliability, especially in highly variable CDR loop regions.

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

Why this matters

Structural characterization of immune receptors—such as antibodies, nanobodies, and TCRs—is critical for understanding antigen recognition, predicting off-target binding, and engineering therapeutics with improved affinity and stability. However, traditional experimental methods are resource-intensive, while general-purpose structure predictors like AlphaFold can take hours to execute for massive datasets of sequences.

ImmuneBuilder resolves this bottleneck by tailoring its neural network architecture to the specific geometrical properties of immune variable domains. By running over 100 times faster than AlphaFold while maintaining comparable CDR-loop accuracy, it empowers researchers to easily integrate structural modeling into high-throughput screening and selection pipelines, facilitating faster and more cost-effective drug discovery and TCR engineering.

• Developed by: Oxford Protein Informatics Group (OPIG)
• Source: Official GitHub repository (oxpig/ImmuneBuilder)
• Reference: ImmuneBuilder Paper (Communications Biology 2023) and TCRBuilder2+ Preprint (2024)