ParaSurf
ParaSurf is a deep learning framework that predicts paratope binding sites by analyzing molecular surfaces of antibodies/nanobodies to identify antigen-binding regions across the entire Fab/Fv domain.
ParaSurf is a deep learning framework that predicts paratope binding sites by analyzing molecular surfaces of antibodies/nanobodies to identify antigen-binding regions across the entire Fab/Fv domain. Combining geometric shape, chemical properties, and electrostatic force fields, ParaSurf employs a hybrid 3D ResNet-Transformer architecture to capture both local and global structural contexts. ParaSurf addresses class imbalance and generalizes to framework residues, enabling antigen-agnostic predictions for therapeutic antibody design and immune interaction studies.
Estimated resource cost: 0.0002945
Categories: Sequence Analysis & Annotation
Tags: Antibodies Binding Site Prediction Proteins
Key Capabilities
- Predicts antibody-binding sites across the entire Fab region, including CDR loops and framework residues, using balanced sampling to mitigate class imbalance.
- Integrates geometric, chemical, and electrostatic surface features (AMBER/CHARMM force fields) for comprehensive interaction modeling.
- Employs a hybrid 3D ResNet-Transformer architecture to capture multi-scale spatial patterns from local atomic environments to global structural contexts.
- Operates antigen-agnostically without requiring partner protein data, and provides open-source code for reproducibility and therapeutic development.
- Achieves state-of-the-art performance on PECAN, Paragraph-expanded, and MIPE benchmarks (AUC-ROC score of 0.973 and AUC-PR score of 0.823).
Runtime Statistics
| Metric | Value |
|---|---|
| runtime_mean | 132 |
| runtime_median | 128 |
| runtime_std | 64 |
| runtime_90th_percentile | 240 |
| runtime_max | 325 |
Similar Tools
- ANARCI
- ANARCII
- CAR-Toner
- DR-BERT
Ready to submit your job?
Review your configuration, then confirm the estimated credit cost before you run the job. Note that credit estimates are not guaranteed and runtime can vary depending on inputs and settings.
Estimated Credits: 0.0002945
Invite-only, limited-time access. Please contact ztang@getantibody.com.
Invite-only, limited-time access. Please contact ztang@getantibody.com.