Model Building of Antibody-Antigen Complex Structures Using GBSA Scores

Area of Science:

• Computational Biology
• Structural Biology
• Biophysics

Background:

• Antibody-antigen complex structure prediction is crucial for antibody drug design and biosensor development.
• Existing scoring functions struggle to accurately assess the stability of protein-protein complexes.

Purpose of the Study:

• To develop an improved method for predicting antibody-antigen complex structures.
• To enhance the accuracy of scoring functions for protein-protein interactions.

Main Methods:

• Utilized the "surFit" docking engine to generate decoy structures for 95 antibody-antigen datasets.
• Developed and applied a generalized Born surface area (GBSA) scoring function combining generalized Born (GB) energy and hydration energy.
• Performed molecular dynamics (MD) simulations on selected decoys to refine predictions.

Main Results:

• The GBSA score achieved a high area under the curve (AUC) of 0.972, outperforming original surFit scores (0.873) and ZRANK scores (0.953).
• MD simulations demonstrated that average GBSA scores from trajectories can effectively reduce non-native decoys with competitive scores.

Conclusions:

• The GBSA scoring function offers a more accurate approach for predicting antibody-antigen complex structures.
• This improved prediction accuracy has significant implications for the rational design of antibody-based therapeutics and diagnostics.