ZYMScott/VRClassification

Nanobody Variable Region Classification Dataset

Dataset Overview

This dataset helps classify the Complementarity Determining Regions (CDRs) in nanobody sequences. CDRs are the highly variable regions in antibodies that form the antigen binding site. Correctly identifying and classifying CDRs is critical for understanding antibody function and structure.

The dataset allows researchers to develop models that can automatically identify CDR regions in nanobody sequences and classify them by type (CDR1, CDR2, or CDR3).

Data Collection

The dataset is based on nanobodies with experimentally determined structures, collected from the Protein Data Bank (PDB) and curated antibody databases. CDR regions were annotated using standard antibody numbering schemes and structural analysis. The dataset includes diverse nanobody sequences with well-defined CDR annotations.

Dataset Structure

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

File Format

CSV files contain these columns:

• seq: Nanobody sequence
• label: class pf nanobody; 0: Frame Region, 1: CDR1, 2: CDR2, 3: CDR3

Uses and Limitations

Uses

• Develop models to automatically identify CDR regions in antibody sequences
• Support antibody structure prediction and engineering
• Help standardize CDR annotation across different antibody databases
• Assist in epitope mapping and paratope analysis

Limitations

• Non-standard CDR definitions may not be represented
• Unusual or engineered nanobodies might have atypical CDR patterns
• Classification is based on sequence features and may not capture structural nuances

Evaluation Metrics

Model performance is evaluated using:

• Accuracy
• F1 score
• Precision
• Recall