Macromolecular Modeling Blog ™

This is the fifth and last post in the CAPRI series, summarizing the presentations of Xiaoqin Zou and Ora Schueler-Furman (Saving the best for last..), as provided by the speakers. I hope the CAPRI series was able to give a snapshot of the state of computational protein-protein docking and its community. I want to thank again to everyone that took part in the meeting and helped me with this series.

A hierarchical approach to protein-protein docking.

Sheng-You Huang & Xiaoqin Zou

Sampling and scoring in protein-protein docking remains challenging because of the difficulty in treating protein flexibility and the expense of an accurate energy function. In this work a hierarchical approach has been developed to predict the binding modes between proteins. As a first step a fast fourier transform (FFT) based docking algorithm is used to globally sample all putative binding modes, in which the protein is represented by a reduced model, that is, each side chain is represented by a single point located at its center of mass. Compared to FFT docking with all-atom models, the FFT docking method with a reduced model is expected to generate more acceptable binding modes because it allows larger side-chain flexibility. Next, the filtered binding modes (normally, several thousands) are refined by an iteratively derived knowledge based scoring function and by considering the flexibility of loops and other backbones. The distance dependant potentials in our scoring function were extracted by a physics based iterative method, which circumvents the long standing reference state problem in the knowledge based approaches. Using the hierarchical protocol we have participated in the CAPRI experiments for rounds 15-19 of 11 targets. In the aspect of prediction we predicted correct binding modes for 4 out of 9 targets. In the aspect of scoring, of the 5 target complexes for which at least one acceptable mode was submitted by the uploaders we obtained correct binding modes for three targets.

Inspection of the performance of the protocol on the CAPRI targets yielded several important conclusions:

• A carefully derived statistical potential-based scoring function benefits both scoring and docking performance.
• Biological information is important to mode selection though does not always help.
• Protein flexibility remains a challenge.

Understanding and modeling of peptide-protein interactions

Barak Raveh*, Nir London*, Dana Movshovitz-Attias and Ora Schueler-Furman

The proper function of living cells is regulated by an intricate network of molecular interactions and signals. A large fraction of these interactions are mediated by short stretches of flexible peptides, which fold upon binding to globular protein receptors. Computational modeling of these interactions has to overcome the complexity involved in simultaneous folding and binding of these molecules, and the large number of degrees of freedom that has to be accounted for. While crude models of peptide-protein interactions may be obtained from homology models, accurate high-resolution models are a necessity for structure-based functional characterization of these interactions, as well as for their controlled manipulation through rational peptide design. As a first step towards this end we have investigated the structural basis of peptide-protein binding strategies.

The main conclusions of this investigation were:

• Most peptides do not induce conformational changes on their partner upon binding
• Peptide-protein interfaces are better packed and contain more hydrogen bonds
• Binding is mediated by peptide hotspots that contribute most of the binding energy
• Peptides tend to bind in the largest pockets or holes on the protein surface

As a second step, we have developed “Rosetta FlexPepDock” a thoroughly tested computational protocol for modeling peptide-protein interactions at high-resolution (Sub-angstrom Modeling of Complexes between Flexible Peptides and Globular Proteins, Proteins (2010) In Press). Starting from approximate structural models of the interactions, the protocol is able is able to account for the considerable diversity of peptide conformations within a given binding site, and to model at near-atomic accuracy residues of conserved binding motifs, which play a key role in cellular regulation, signaling and disease.

Rosetta FlexPepDock: Sub-angstrom Modeling of Complexes between Flexible Peptides and Globular Proteins

Finally, we analysed peptide-protein interactions in the context of protein-protein interactions. Specifically, we asked what part of protein-protein interactions are mediated largely by a single linear peptide stretch from one of the partners. This analysis showed that a large portions of protein-protein interactions (from the CAPRI and Docking Benchmark 3.0 datasets) are indeed mediated by a single peptide contributing most of the interaction energy. This strategy is often used in specific Enzyme-Inhibitor interactions. Furthermore we examined the distribution of RMSDs to the free form of the same peptide stretch (within the unbound monomer) and show that for peptides that undergo significant conformational change, FlexPepDock is able to recover the native binding mode. We believe that these studies bring us closer towards the ultimate goal of manipulating peptide-protein interactions.

Raveh B*, London N* & Schueler-Furman O (2010). Sub-angstrom Modeling of Complexes between Flexible Peptides and Globular Proteins. Proteins. In Press.

London N, Movshovitz-Attias D, & Schueler-Furman O (2010). The Structural Basis of Peptide-Protein Binding Strategies. Structure (London, England : 1993), 18 (2), 188-199 PMID: 20159464