The scoring of poses in protein-protein docking: current capabilities and future directions

Abstract

BackgroundProtein-protein docking, which aims to predict the structure of a protein-protein complex from its unbound components, remains an unresolved challenge in structural bioinformatics. An important step is the ranking of docked poses using a scoring function, for which many methods have been developed. There is a need to explore the differences and commonalities of these methods with each other, as well as with functions developed in the fields of molecular dynamics and homology modelling.ResultsWe present an evaluation of 115 scoring functions on an unbound docking decoy benchmark covering 118 complexes for which a near-native solution can be found, yielding top 10 success rates of up to 58%. Hierarchical clustering is performed, so as to group together functions which identify near-natives in similar subsets of complexes. Three set theoretic approaches are used to identify pairs of scoring functions capable of correctly scoring different complexes. This shows that functions in different clusters capture different aspects of binding and are likely to work together synergistically.ConclusionsAll functions designed specifically for docking perform well, indicating that functions are transferable between sampling methods. We also identify promising methods from the field of homology modelling. Further, differential success rates by docking difficulty and solution quality suggest a need for flexibility-dependent scoring. Investigating pairs of scoring functions, the set theoretic measures identify known scoring strategies as well as a number of novel approaches, indicating promising augmentations of traditional scoring methods. Such augmentation and parameter combination strategies are discussed in the context of the learning-to-rank paradigm.

CODV-Ig, a universal bispecific tetravalent and multifunctional immunoglobulin format for medical applications

“…Protein-protein docking is revealed as a sufficiently robust method to identify suitable domain arrangements for protein engineering, even if it is not reliable enough to routinely and unambiguously predict protein complexes or domain configurations as they are experimentally observed. 66,67 Indeed, the modeling approach proved surprisingly efficient as we experimentally explored only »0.08% of the search space and identified almost exclusively constructs with expected molecular weight that maintain parental target affinities.…”

Integrative computational modeling of protein interactions

“…Another common strategy in scoring is to use knowledge-based functions derived from statistical analysis of experimentally determined structures. A recent study [47] identified several knowledge-based scoring functions that seem to discriminate near-native models particularly well: SIPPER [48], DECK [49] and PISA/PIE [50]. In addition to these recently developed functions, others such as ZRANK [51] and DFIRE [52] have been in use for years in the docking community.…”

“…DECK [49], DCOM-PLEX [52], PIE/PISA [50] and SPIDER [53]). Also, the combination of scores from different functions is not necessarily additive, given the redundancy of their individual terms, and requires special attention and expert knowledge [47].…”

Analysis and Ranking of Protein-Protein Docking Models Using Inter-Residue Contacts and Inter-Molecular Contact Maps

“…Although the authors of the E2-inhibitor structure were right in predicting that the second inhibitor molecule was exploring an additional region on the E1-E2 interaction, Figure 6b clearly shows that most of it still remains available for possible inhibitor binding. The E2 region unaffected by the inhibitor binding, rimmed in green in Figure 6b and shadowed in green in the corresponding contact map, involves in particular the C-terminal part of helix A (residues [16][17][18][19][20][21][22][23][24][25][26] and the loop connecting helix B to helix C (residues 61-68). This region could be possibly exploited for the design of more potent inhibitors of the E1-E2 interaction.…”

“…However, a dramatic disproportion still exists between the number of experimental structures solved for protein complexes and the number of structures available for single proteins [11]. In this scenario, the prediction of a protein-protein complex structure by molecular docking and the scoring of obtained models (topics recently reviewed in [12][13][14][15][16], respectively) will most probably assume a more and more relevant role in drug design.…”