Computational Modeling of Protein Interactions

Protein–protein interactions were predicted using the STRING functional protein association network [66 (link)]. Coordinate files for homology models were determined using sequence identity criteria (average identity 37%). The DSS1 model structures were created using the Phyre2 protein folding recognition server by uploading the following sequences: AtDSS1(I) (Uniprot:AT1G64750) and AtDSS1(V) (Uniprot:AT5G45010) [67 (link)]. The best template structure, human 26S proteasome bound to ubiquitin carboxyl-terminal hydrolase 14 (USP14)-ubiquitin-like modifier-activating enzyme 1 (UbAl) (pdb:5gjg) has 66% identity with a query sequence [68 (link)]. After preparing the protein data bank (PDB) model, the docking protocol was applied using the PatchDock algorithm with a cluster root mean square deviation (RMSD) of 4 Å [69 (link)]. The top 20 cluster structures were subjected to FireDock refinement based on transformations [70 (link)]. The scoring function is based on geometric fitting and atomic desolvation energy [71 (link)]. The parameters used to calculate the solution global binding energy (GBE) are the atomic contact energy or desolvation energy for the two proteins at transit from the unbound state to the complex (ACE), hydrogen and disulfide bonds (HB) and aliphatic interactions (ALIPH), attractive and repulsive van der Walls forces, short- and long-range Coulomb forces, cation-π and π-π stacking interactions.