Graph-theoretical comparison of protein surfaces reveals potential determinants of cross-reactivity and the molecular mimicry.

Different proteins, even without sequence similarity, still can contain similar surface regions involved in protein-protein interactions with common target. These regions can serve as structural determinants of cross-reactivity and molecular mimicry. Molecular mimicry, defined as the process in which structural properties of one molecule are simulated by the dissimilar molecules, is implicated in several biologically important processes, including autoimmune and allergic reactions, binding of some ligands to common receptor, and interactions in cell signaling. The problem of identification of the determinants of molecular mimicry is not completely solved at this time. We hypothesize that identification of structurally and chemically similar surface regions of two protein molecules capable of binding to the same target will allow us to identify sites involved in cross-reactivity including determinants of the molecular mimicry. We used a graph-theoretical approach in order to determine highly similar surface regions of two proteins with known three-dimensional structures. This approach uses a variation of Maximal Common Subgraph (MCS) isomorphism, where an association graph is constructed based on the surface-exposed residues of the two molecules and the matching regions are found based on the maximum cliques in the association graph. Testing the proposed method on the targets of autoantibody involved in antiphopholipid syndrome (APS)--beta2-GPI, PC, thrombin, factor IX, factor X, and plasmin allowed identifying potential epitopes for antibody that can inhibit coagulation proteases. Application of this method to the Activated Protein C and factor VII Gla-domains revealed surface regions involved in EPCR and plasma membrane binding, consistent with known experimental results. Analysis of major pollen allergen that can cause food allergies through cross-reactivity found known epitopes involved in cross-reactivity and also revealed additional surface regions that can complement the list of epitopes. Taken together, our results suggest that the proposed graph-theoretical approach can identify determinants of cross-reactivity and molecular mimicry.