Nanobody Paratope Prediction Dataset

Dataset Overview

This dataset helps predict which amino acids in nanobody sequences directly bind to antigens (these binding sites are called "paratopes"). Knowing the paratope residues is important for understanding how antibodies interact with their targets and for designing better therapeutic nanobodies.

Data Collection

The data is based on solved 3D structures of nanobody-antigen complexes. These structures come from the Protein Data Bank (PDB) and scientific literature. We identified paratope residues by analyzing the contact interfaces between nanobodies and antigens.

Dataset Structure

The dataset is split into training, validation, and test sets.

File Format

CSV files contain these columns:

• seq_nanobody: Nanobody amino acid sequence
• seq_antigen: Antigen amino acid sequence
• paratope: Binary list indicating if each residue is part of the paratope (1) or not (0)

antigen_embeddings.pt: precomputed antigen sequence embeddings through ESM-2 (650M)

Uses and Limitations

Uses

• Develop models to predict nanobody binding sites
• Help design and engineer nanobodies
• Understand nanobody-antigen binding mechanisms

Limitations

• Limited number of experimentally solved nanobody-antigen structures
• Paratope definitions may vary depending on experimental conditions
• Different antigen types may cause different paratope patterns

Evaluation Metrics

Model performance is evaluated using:

• Accuracy
• F1 Score
• Precision
• Recall
• AUROC (Area Under the Receiver Operating Characteristic curve)
• AUPRC (Area Under the Precision-Recall Curve)