Superposition of Cryo-EM and AlphaFold Predictions of Dengue Antigen-Antibody Complexes

Authors: Chen J*

DOI: 10.23880/ijbp-16000270

Volume 11, Issue 1

Abstract

Accurate prediction of antibody-antigen interfaces is essential for structure-based drug design (SBDD) an vaccine development. The dengue virus envelope (E) protein is a major target of neutralizing antibodies and has been extensively characterized structurally. Although the deep-learning methods such as AlphaFold has achieved near-experimental accuracy for many protein structure which has strong MSA support, the reliability of the model for predicting the antibody-antigen complexes has not been sufficiently assessed. Here, we used 5 independent predictions from AlphaFold 3 to compare with a high-resolution cryo-EM structure of the dengue virus serotype 2 (DENV-2) E protein in complex with neutralizing antibody 2D22 (PDB: 8Y3H). AlphaFold 3 was able to reproduce the global fold of the E protein precisely; however, there was significant heterogeneity observed at the antibody-antigen interface. The predicted complexes were clustered into three separate binding modes, all three of which did not contain the experimentally determined epitope centered around residues A150 - 154 and A360 - 364. The predicted interfaces did not have the exact residue-level interactions that were seen in the cryo-EM structure. These results show that although AlphaFold 3 can reliably predict the backbone structure of antigens, it is still constrained in precisely positioning the location of the binding site of the antibody against flexible viral antigens. We show the importance of experimental proof and integrative modeling over the computational prediction in using them for guiding antiviral antibodies and vaccines design.

Keywords: Dengue Virus; Alphafold; Cryo-Electron Microscopy (Cryo-EM); Antibody-Antigen Complex, Epitope Mapping

View PDF