A new approach for extracting network information from protein dynamics

Increased knowledge about protein structures has increased interest in molecular dynamics simulations to study function. To analyze these simulations, proteins are modeled as networks to take advantage of well-developed methods from network science. Protein networks are often constructed from correlation measures. Yet, it is clear in network science that solving the inverse problem reconstructs network interactions. Thus, we apply this inverse approach to the dynamics of protein dihedral angles, a system of internal coordinates that avoids the structural alignment artifacts. Using the well-characterized adhesion protein, FimH, we show that our method identifies networks that are physically interpretable and robust. In FimH, we detect differences in inferred networks consistent with the allosteric pathway sites. Next, we use our approach to detect dynamical differences, despite structural similarity, for two other adhesion proteins: Siglec-8 in the immune system and the SARS-CoV-2 spike protein. Our approach enables us, for example, to validate a new mechanism to explain the stability of a Siglec-8 binding pocket loop. Thus, using an inverse approach to extract a protein network makes it tractable to apply analysis techniques, e.g. community detection, without edge pruning.