Nanobody Type Classification Dataset

Dataset Overview

This dataset helps classify different types of single-domain antibodies (sDAbs). Nanobodies are a special type of single-domain antibody, mainly from camelid heavy-chain antibodies. Correctly identifying different types of sDAbs is important for understanding their structural properties, binding ability, and potential applications.

The dataset contains sDAb sequences from different sources, with the goal of classifying them into different types, such as VHH (camelid antibodies), VH (traditional antibody heavy chains), and others.

Data Collection

The sequences in this dataset come from multiple public databases, including the Protein Data Bank (PDB), Antibody Database (AbDb), and published literature. All sequences have been verified by experts to ensure classification accuracy.

Dataset Structure

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

File Format

CSV files contain these columns:

• seq: Single-domain antibody amino acid sequence
• label: Antibody type classification (category label); 0: VHH from llama, 1: VH from human, 2: VNAR from shark, 3: V lambda from human, 4: V kappa from human.

Uses and Limitations

Uses

• Develop models to accurately classify sDAb types
• Help researchers quickly identify the type of newly discovered sequences
• Provide references for antibody engineering and design

Limitations

• Some antibody types may be underrepresented in the dataset
• Non-natural designed antibodies may be difficult to classify accurately
• Structural and sequence similarities between types may present classification challenges

Evaluation Metrics

Model performance is evaluated using:

• Accuracy
• F1 Score
• Precision
• Recall