In Silico Investigation of Molecular Mimicry in Antibody-Antigen Complexes: A Structural Prediction Pipeline for Targets Lacking Experimental Structures

A single antibody can occasionally bind targets with no evolutionary relationship, a phenomenon termed molecular mimicry. The neutralizing anti-SARS-CoV-2 antibody CV07-200 recognizes both the viral Spike receptor-binding domain (RBD) and RSSL-01370, a surface protein of the commensal bacterium Streptococcus salivarius, despite no detectable sequence or structural homology. This thesis investigates the structural basis of that cross-reactivity through an integrative computational pipeline for targets lacking experimental structures. BLASTp and Dali superposition (Z = 0.1) first excluded sequence and fold homology, shifting the analysis to the molecular surface. MaSIF-derived surface fingerprinting, encoding electrostatics, hydrogen bonding, hydrophobicity, and shape, identified a bacterial patch whose biochemical profile matches the viral epitope well beyond chance (Z = 3.03). Docking and explicit-solvent molecular dynamics then showed that CV07-200 engages both antigens through the same aromatic-and-polar paratope core, while each presents a distinct yet complementary surface: an aromatic "warhead" on the RBD and an acidic "trap" on RSSL-01370, both stabilized predominantly by van der Waals and aromatic packing. These findings support a model in which molecular mimicry arises from convergent surface biochemistry rather than shared sequence or fold, and point toward surface-based conformational epitope prediction.