NABP-BERT: NANOBODY®-antigen binding prediction based on bidirectional encoder representations from transformers (BERT) architecture
2025

NABP-BERT: Predicting Nanobody-Antigen Binding Using Deep Learning

Sample size: 1314 publication 10 minutes Evidence: high

Author Information

Author(s): Ahmed Fatma S, Aly Saleh, Liu Xiangrong

Primary Institution: Xiamen University

Hypothesis

Can a deep learning model based on BERT architecture effectively predict nanobody-antigen binding from sequence information?

Conclusion

The NABP-BERT model outperforms existing methods in predicting nanobody-antigen interactions.

Supporting Evidence

  • NABP-BERT achieved an AUROC of 0.986 and an AUPR of 0.985.
  • The model was trained on a dataset of 1314 nanobody-antigen pairs.
  • Pretraining on a large dataset improved the model's performance significantly.
  • NABP-BERT outperformed traditional machine learning methods in binding prediction.

Takeaway

This study created a computer program that can guess how well tiny antibodies stick to germs, which can help make better medicines.

Methodology

The model was trained using three distinct databases, focusing on nanobody-antigen and protein-protein interaction datasets.

Potential Biases

Potential biases in training data could affect the model's predictions.

Limitations

The model's performance may vary with different configurations of attention heads and encoder layers.

Statistical Information

P-Value

0.986

Statistical Significance

p<0.05

Digital Object Identifier (DOI)

10.1093/bib/bbae518

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication