Figure 2.

Flowcharts for the three representative template-based complex structure prediction strategies. (a) Dimeric threading method. The black lines outline a threading procedure, similar to MULTIPROSPECTOR [24], which identifies complex templates from a dimer template library by dimeric query-to-template alignments. Blue lines indicate additional steps that improve upon the base method by utilizing a monomer template library and structural superposition, similar to COTH [18•]. Parts in magenta indicate stages where interface evaluation is used to increase alignment accuracy, ranking, and specificity. (b) Monomer threading and oligomer mapping. The protocol was used in SPRING [21••] where a combined template library containing both monomer and oligomer proteins is used. Monomeric threading is first used to identify a list of templates for each monomer chain where some templates will be parts of oligomers. The complex models are constructed by mapping the top templates of each monomer onto the framework excised from the associated oligomers, and ranked by monomer threading and interface matching scores. (c) Template-based docking. In this protocol, full-length models or experimental structures of the monomer proteins are matched against the dimer template library based on either global fold or interface structure comparisons. Dimer templates are selected from the complexes which have both components structurally similar to monomer structure of the target chains. A similar protocol is used in PrePPI [20••], PRISM [23] and the approach by Vakser et al [19•,22•].