SARS-CoV-2 Nanobody Interaction Dataset (AVIDa-SARS-CoV-2)

Dataset Overview

This dataset features the antigen-variable domain of heavy chain antibody (VHH) interactions obtained from two alpacas immunized with SARS-CoV-2 spike proteins. It includes binary labels indicating the binding or non-binding of diverse VHH sequences to 12 SARS-CoV-2 mutants, including variants such as Delta and Omicron.

The dataset is designed to evaluate the ability of computational models to predict nanobody binding across different viral variants, which is crucial for understanding antibody cross-reactivity and developing therapeutics against emerging variants.

Data Collection

The data was collected from experiments with two alpacas immunized with SARS-CoV-2 spike proteins. Binding assays were conducted to determine whether specific VHH sequences bind to different SARS-CoV-2 variants. Further details are described in the paper "A SARS-CoV-2 Interaction Dataset and VHH Sequence Corpus for Antibody Language Models."

Dataset Structure

The dataset is strategically split into training, validation, and test sets to evaluate model generalization:

• Training set: Covers wild-type (WT), mild mutations (D614G), conserved regions (S2-domain), and immune escape variants (PMS)
• Validation set: Contains Kappa and Lambda variants that models have never seen before, useful for hyperparameter tuning
• Test set: Includes Alpha, Beta, Delta, and Omicron variants for final real-world generalization assessment

File Format

Main Dataset Files (train.csv / val.csv / test.csv)

Antigen Sequences embeddings (antigen_embeddings.pt)

precomputed antigen sequence embeddings through ESM-2 (650M)

Uses and Limitations

Uses

• Develop models to predict nanobody binding to SARS-CoV-2 variants
• Identify nanobodies with broad cross-reactivity against multiple variants
• Understand the impact of viral mutations on antibody recognition
• Design therapeutic nanobodies with resistance to viral escape

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)