Use ImmuneBuilder Online
Commercially Available ImmuneBuilder No-Code Web Server
Try ImmuneBuilder
ImmuneBuilder: High-Speed Structural Modeling for the Adaptive Immune System
ImmuneBuilder is a specialized suite of deep learning models designed for the high-throughput, high-accuracy structural prediction of antibodies, Nanobodies, and T-cell receptors (TCRs). Developed by the Oxford Protein Informatics Group (OPIG), ImmuneBuilder addresses the unique structural challenges of immune receptors—specifically the hypervariable loops that govern binding—with a speed that is 100 times faster than general-purpose tools like AlphaFold2.
By utilizing an architecture that predicts the distance and orientation between residues, ImmuneBuilder enables the structural characterization of entire immune repertoires, providing a robust foundation for therapeutic discovery and vaccine design.
Key Innovations: Specialized Geometric Learning
ImmuneBuilder moves beyond general protein folding by utilizing an ensemble of deep learning models trained specifically on the structural geometry of the immune system.
Modality-Specific Models: Features three specialized pipelines: ABodyBuilder2 for antibodies, NanoBuilder for Nanobodies, and TCRBuilder2 for T-cell receptors.
Ensemble Confidence Scoring: Employs an ensemble of four independent models for each modality, allowing the system to provide per-residue confidence estimates ($RMSD$) based on the variance between model predictions.
Geometric Constraint Satisfaction: Predicts inter-residue distances and orientations (Cβ and N - O distances) to reconstruct all-atom structures that adhere to strict biological physics.
High-Throughput Repertoire Mining: Capable of generating thousands of accurate structural models in minutes, enabling the analysis of vast B-cell and T-cell repertoire data.
Open-Source & Lightweight: Unlike resource-heavy models, ImmuneBuilder is designed to run efficiently on standard hardware while maintaining competitive accuracy.
Performance Benchmarks
ImmuneBuilder consistently outperforms general-purpose structural predictors and traditional homology-based methods in both speed and loop accuracy.
Task | Generalist Method (AF2) | Homology Modeling | ImmuneBuilder Result | Key Finding |
CDR-H3 Accuracy | 3.1 Å | 3.4 Å | 2.8 Å | Superior precision on hypervariable loops |
Prediction Speed | Minutes per sequence | Hours (with MD) | Seconds | 100x faster than AlphaFold2 |
Nanobody Folding | Moderate | Poor | High Accuracy | Specialized for single-domain receptors |
TCR CDR3 Loop | 1.9 Å | 2.3 Å | 1.6 Å | SOTA performance for T-cell receptors |
Scientific Breakthroughs in Immune Discovery
Repertoire-Scale Structural Characterization
Traditional structural biology is limited by experimental throughput. ImmuneBuilder allows researchers to model the "structural dark matter" of the adaptive immune system, transforming millions of unannotated sequences into 3D models to identify novel binding motifs and structural clusters across whole populations.
Nanobody & TCR Engineering
Nanobodies and TCRs present unique modeling challenges due to their single-domain nature or specific loop orientations. NanoBuilder and TCRBuilder2 are specifically tuned for these geometries, providing accurate paratope predictions that are critical for developing next-generation biologics and CAR-T therapies.
Automated Confidence Filtering
By utilizing its ensemble architecture, ImmuneBuilder can automatically flag "hard-to-model" sequences where model variance is high. This allows researchers to focus experimental resources on high-confidence leads while identifying structural outliers that may require further optimization.
ImmuneBuilder on Tamarind Bio: Repertoire Analysis at Scale
Tamarind Bio provides a managed, high-performance environment to execute ImmuneBuilder’s repertoire-scale workflows without the need for complex software installation or GPU management.
No-Code Dashboard: Launch ABodyBuilder2, NanoBuilder, or TCRBuilder2 jobs through a simple web interface.
Automated Batch Processing: Upload CSV or FASTA files containing thousands of sequences and receive a complete structural library in minutes.
How to Use ImmuneBuilder on Tamarind Bio
Access the Toolkit: Log in to tamarind.bio and select the ImmuneBuilder tool for your target modality (Antibody, Nanobody, or TCR).
Input Sequences: Provide your variable region sequences. For antibodies, enter both Heavy and Light chain sequences for joint modeling.
Choose Modeling Engine: Select the appropriate engine—ABodyBuilder2, NanoBuilder, or TCRBuilder2.
Run Structural Repertoires: The platform executes the ensemble models to predict distances, orientations, and 3D coordinates.
Evaluate Confidence Scores: Review the per-residue RMSD estimates to identify the most reliable structural regions, particularly in the CDR loops.
Export & Analyze: Download high-resolution PDB files and structural reports for downstream docking, epitope mapping, or developability screening.