EvoNB: A protein language model-based workflow for nanobody mutation prediction and optimization.

Area of Science:

• Biotechnology
• Computational Biology
• Protein Engineering

Background:

• Optimizing nanobodies for therapeutic use is vital but often complex and time-consuming.
• Current methods for identifying beneficial nanobody mutations can be inefficient, limiting their practical application.

Purpose of the Study:

• To develop an efficient workflow for predicting and optimizing nanobody mutations.
• To enhance the therapeutic potential of nanobodies by improving their binding affinity to target antigens.

Main Methods:

• Developed the Evolutionary-Nanobody (EvoNB) workflow, integrating protein language models (PLMs) and molecular dynamics (MD) simulations.
• Fine-tuned the ESM2 protein language model on a large nanobody dataset for improved sequence feature capture.
• Validated predicted mutations using MD simulations on representative nanobody-antigen complexes to assess binding affinity changes.

Main Results:

• The EvoNB workflow demonstrated enhanced predictive accuracy for nanobody mutations, particularly in conserved and variable regions.
• MD simulations confirmed that EvoNB-identified mutations significantly improved nanobody-antigen binding affinity.
• The sequence-based prediction approach is less reliant on structural data and integrates well with tools like AlphaFold 3.

Conclusions:

• The EvoNB workflow offers an effective tool for the rapid design and optimization of nanobody mutations.
• This approach accelerates the identification of promising nanobody variants for experimental validation, bypassing traditional evolutionary methods.
• EvoNB facilitates the broader application of PLMs in nanobody engineering and the biomedical field.