The overall procedure of the method. The flowchart of systematically constructing and testing the interface fragment pair library is shown in (a). We first collected all pairs of interface fragments from the 3did database. These data were then clustered. A subset of recurrent structures were further selected as the library. Finally, the library was used to guide complex assembly. During the complex assembly, we assigned a sliding window along each protein domain. The Cα atoms in the windows were aligned to each fragment pair in the library (b). If RMSDs are smaller than the cutoff value, we further superimposed the structures of entire domains to their corresponding fragments according to the relative position of windows in each domain. After both domains were aligned to the interface fragment pair, a dimeric complex was constructed.