De novo generation of SARS-CoV-2 antibody CDRH3 with a pre-trained generative large language model

Pre-trained deep protein models have become essential tools in fields such as biomedical research, enzyme engineering, and therapeutics due to their ability to predict and optimize protein properties effectively. However, the diverse and broad training data used to enhance the generalizability of these models may also inadvertently introduce ethical risks and pose biosafety concerns, such as the enhancement of harmful viral properties like transmissibility or drug resistance. To address this issue, we introduce a novel approach using knowledge unlearning to selectively remove virus-related knowledge while retaining other useful capabilities. We propose a learning scheme, PROEDIT, for editing a pre-trained protein language model toward safe and responsible mutation effect prediction. Extensive validation on open benchmarks demonstrates that PROEDIT significantly reduces the model's ability to enhance the properties of virus mutants without compromising its performance on non-virus proteins. As the first thorough exploration of safety issues in deep learning solutions for protein engineering, this study provides a foundational step toward ethical and responsible AI in biology.

show abstract

Deep generative design of neutralizing nanobodies against SARS-CoV-2 variants

Huo,

Tian,

et al. 2024

Preprint

View full text Add to dashboard Cite

In recent years, single-domain heavy chain antibodies (nanobodies), with only one-tenth the molecular weight of conventional antibodies, have emerged as important therapeutic proteins in the fight against SARS-CoV-2. However, the rapid mutation of the virus often renders existing nanobodies ineffective, underscoring the need to develop nanobodies that specifically target new variants. Traditional methods for discovering nanobodies are time-consuming and complex, making it difficult to efficiently identify nanobodies that bind to specific epitopes. To address this, we propose a de novo nanobody design method based on a Generative Adversarial Network (GAN). We developed a deep generative model, AiCDR, which consists of three discriminators and one generator. By determining the contribution of these three discriminators to the entire network, we can enhance the discrimination of generated sequences and reduce the similarity of the generated CDR3 sequences with natural peptides and random sequences, thereby ensuring their nature-like properties. These generated CDR3 sequences were then grafted onto humanized nanobody scaffolds, resulting in a structural library of approximately 104 nanobodies with natural-like properties. Using computational methods, we screened this library against the Spike (S) protein of the SARS-CoV-2 Omicron variant and identified 10 candidate nanobodies. Functional assays confirmed that two of these nanobodies exhibited neutralizing activity against the S protein. Our study demonstrates the potential of deep learning in nanobody design and offers a novel approach for developing nanobodies that target rapidly evolving viral variants.

show abstract

De novo generation of SARS-CoV-2 antibody CDRH3 with a pre-trained generative large language model

Cited by 5 publications

References 70 publications

A deep drug prediction framework for viral infectious diseases using an optimizer-based ensemble of convolutional neural network: COVID-19 as a case study

A deep drug prediction framework for viral infectious diseases using an optimizer-based ensemble of convolutional neural network: COVID-19 as a case study

Unlearning Virus Knowledge Toward Safe and Responsible Mutation Effect Predictions

Deep generative design of neutralizing nanobodies against SARS-CoV-2 variants

Contact Info

Product

Resources

About