How to accurately predict nanobody structure: Classical physics-based simulations or deep learning approaches.

Area of Science:

• Structural biology
• Immunology
• Computational biology

Background:

• Antibodies are crucial proteins for disease prevention, diagnosis, and treatment.
• Camel heavy-chain single-domain antibodies (VHHs), or nanobodies (Nbs), offer advantages over full-length antibodies due to their small size, stability, and affinity.
• Accurate prediction of nanobody complementarity-determining regions (CDRs), particularly CDR3, is essential for understanding antigen binding but remains a significant challenge.

Purpose of the Study:

• To systematically evaluate structure prediction strategies for different categories of nanobodies (Nbs).
• To assess the accuracy of CDR3 structure prediction using physics-based simulations and deep learning methods.
• To provide insights into the mechanism of antigen binding by Nbs.

Main Methods:

• Selected representative Nbs (Nb32, Nb80, Nb35) from concave, loop, and convex categories with known structures.
• Employed physics-based simulations, including homology modeling and molecular dynamics simulations.
• Utilized deep learning models, specifically AlphaFold2 and RoseTTAFold, for structure prediction.

Main Results:

• Compared predicted Nb structures with experimental data to evaluate prediction accuracy, focusing on CDR3.
• Identified varying prediction accuracies across different Nb categories and prediction strategies.
• The study suggests that nanobody-target protein binding likely occurs via an induced fit mechanism.

Conclusions:

• Developed suggestions for improving the accurate prediction of nanobody structures across different categories.
• The findings contribute to a better understanding of nanobody structural dynamics and antigen-binding mechanisms.
• Highlights the potential of combining computational methods for enhanced nanobody structure prediction.