Epitope profiling of coronavirus-binding antibodies using computational structural modelling

Abstract

Identifying the epitope of an antibody is a key step in understanding its function and its potential as a therapeutic. It is well-established in the literature that sequence-based clonal clustering can identify antibodies with similar epitope complementarity. However, there is growing evidence that antibodies from markedly different lineages but with similar structures can engage the same epitope with near-identical binding modes. Here, we describe a novel computational method for epitope profiling based on structural modelling and clustering, and show how it can identify sequence-dissimilar antibodies that engage the same epitope. We start by searching for evidence of structural conservation across the latest solved SARS-CoV-2-binding antibody crystal structures. Despite the relatively small number of solved structures, we find numerous examples of sequence-diverse but structurally-similar coronavirus-binding antibodies engaging the same epitope. We therefore developed a high-throughput structural modeling and clustering method to identify functionally-similar antibodies across the set of thousands of coronavirus-binding antibody sequences in the Coronavirus Antibody Database (CoV-AbDab). In the resulting multiple-occupancy structural clusters, 92% bind to consistent domains based on CoV-AbDab metadata. Our approach functionally links antibodies with distinct genetic lineages, species origins, and coronavirus specificities. This indicates greater convergence exists in the immune responses to coronaviruses than would be suggested by sequence-based approaches. Our results show that applying structural analytics to large class-specific antibody databases will enable high confidence structure-function relationships to be drawn, yielding new opportunities to identify functional convergence hitherto missed by sequence-only analysis.

Unconstrained generation of synthetic antibody–antigen structures to guide machine learning methodology for antibody specificity prediction

“…creates a discretized lattice representation of the protein antigen (Figure 1B), by minimizing the dRMSD between the original PDB structure and its many possible lattice counterparts 64 (Supplementary Figure 3A, Methods). Protein glycosylation is modeled as "inaccessible positions" on the surface of the discretized antigen (Supplementary Figure 3A,B), which impact the binding affinity of the CDRH3 sequences (Supplementary Figure 4G), as observed experimentally 65,66 . We optimized the lattice resolution (distance between neighboring amino acids) to reach an average RMSD of 3.5Å to the original PDB (Supplementary Figure 3C-F).…”

Current advances in biopharmaceutical informatics: guidelines, impact and challenges in the computational developability assessment of antibody therapeutics

“…Incorporating structural information from models can allow prediction of antibody properties in a repertoire and we may be able to predict antibody domain binding by performing structural clustering of antibody models with known-function antibody datasets, such as CoV-AbDab. 162 , 163 …”

Deciphering the language of antibodies using self-supervised learning

“…Most work focuses on analysing the third CDR (CDR3) of the BCR heavy chain, as it is the greatest determinant of binding; however, predicting CDR3 structure and function is notoriously difficult (11)(12)(13). Sequence-dissimilar CDR3s can adopt similar structures (14) and recognise similar regions of a target molecule (15), while small changes in CDR3 sequence can change structure and binding properties (16,17). In addition, BCRs with identical CDR3 sequences but changes elsewhere can have different binding properties (18,19).…”