A Unified Dataset for Antibody and Nanobody Design Including Sequence, Structure, and Binding Affinity Data

Area of Science:

• Biotechnology
• Computational Biology
• Immunology

Background:

• Deep generative models show promise for antibody and nanobody design.
• Existing datasets are fragmented and inconsistent, limiting model training.
• There is a need for a unified, comprehensive dataset for antibody and nanobody research.

Purpose of the Study:

• To introduce the Antibody and Nanobody Design Dataset (ANDD).
• To create a unified resource integrating sequence, structure, antigen, and affinity data.
• To facilitate the training of deep generative models for antibody and nanobody design.

Main Methods:

• Integrated data from 15 diverse sources into a single dataset.
• Included antibody/nanobody sequences, structural data, and antigen sequences.
• Augmented affinity data using the ANTIPASTI model for binding affinity prediction.

Main Results:

• ANDD comprises 48,683 antibody/nanobody sequences and 24,941 structural entries.
• The dataset contains 12,575 antigen sequences and 9,557 affinity values.
• ANDD is the largest dataset for antibody/nanobody-antigen pairs with affinity data.

Conclusions:

• ANDD addresses data fragmentation and inconsistency challenges.
• The dataset provides a robust foundation for training deep generative models.
• ANDD enables better modeling of antibody/nanobody-antigen interactions for novel therapeutic development.