Nanobody (VHH) Affinity Prediction Dataset

Dataset Overview

This dataset helps predict the binding affinity between nanobodies (VHH, single-domain antibodies from camelids) and their target antigens. Affinity is a key parameter that measures how strongly an antibody binds to its antigen, usually expressed as dissociation constant (KD) or binding free energy.

High affinity is a critical property for therapeutic antibodies, so accurately predicting nanobody affinity is important for antibody engineering and screening.

Data Collection

The dataset is based on experimentally measured nanobody-antigen binding affinities. Data is collected from published literature and split based on clustering (sequence similarity).

Dataset Structure

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

File Format

CSV files contain these columns:

• seq: Nanobody amino acid sequence
• score: Affinity value (typically -log10(KD) where KD is in M), higher values indicate stronger binding affinity

Uses and Limitations

Uses

• Develop models to predict nanobody affinity
• Help select and optimize nanobodies
• Reduce experimental work and speed up drug development

Limitations

• Differences in affinity measurement methods may cause data variability
• The same antibody-antigen pair may have different affinity values under different conditions
• The dataset may not cover all possible nanobody-antigen combinations

Evaluation Metrics

Model performance is evaluated using:

• Spearman correlation
• R²
• Root Mean Squared Error (RMSE)
• Mean Absolute Error (MAE)