Unit: Structural Bioinformatics
Director: Michael Nilges
The main research interests of the unité are the analysis and prediction of the structure and interactions of proteins, the prediction of large motions of proteins, the determination of structure and dynamics of proteins from NMR data, and the analysis of genomic data. The scope is from atomic detail to the analysis of whole genomes. Methods used and/ or developed include homology modelling, docking, molecular dynamics, and free energy calculations.
Analysis of protein interactions
Modelling of protein-peptide interactions in snake venoms (R. Maroun)
The venom of the snake B.jararaca contains two serine proteinases that show 66% sequence identity, but present complementary fonctions. PA-BJ induces platelet aggregation without clotting fibrinogen and bothrombin clotts fibrinogen, but has no direct effects on platelets. In order to understand the molecular basis of the distribution of the 2 major procoagulant functions of thrombin -fibrinogen activation and platelet aggregation- among the 2 snake venom enzymes, we obtained molecular models of the enzymes in complex with their natural ligands. For bothrombin, the ligand at the active site and at exosite I is a fragment of fibrinogen containing fibrinopeptide A. PA-BJ is in the presence of a peptide fragment coming from the binding site of thrombin on the thrombin receptor, PAR-1. The molecular models of the complexes thus obtained show, in an analogous fashion to thrombin, that each snake venom enzyme possesses 2 exosites. Nevertheless, the exosites contain a lesser proportion of basic amino acids than thrombin (45-72%), which could explain the reduction in the functional diversity of these enzymes. Besides, the composition of exosite I is different for each of the 2 enzymes. We have identified those residues in exosite I that could contribute to the differences in specificity. Finally, as opposed to thrombin, allostery does not seem to play a role in the macromolecular substrate recognition by these enzymes.
New free energy methods to analyze protein-ligand associations (A. Blondel)
The aim of these developments is to make the quantitative predictions of the free energy of protein-ligand association sufficiently reliable to be a useful complement to experiments on biological processes or to contribute in the design of new drugs. Because of their energetic importance, we are also interested in the identification of the relevant conformations and mechanisms for protein-ligand association or for more complex processes such as enzymatic reactions or folding.
The conception of these methods was motivated by a thorough study of the association of R67 DHFR subunits combining experimental and computational approaches (see previous reports, Protein Folding and Modelling Unit). Our new method reduces significantly the uncertainties of free energy difference calculations. The method is based on the optimization of ensemble fluctuations during thermodynamic integration. Good convergence properties were observed, and predictions were in good agreement with experiments (~0.4 kcal/mol difference). A recent comparison with standard methods from the literature confirmed the superiority of our method. We further tested the convergence on systems presenting significant hysteresis. Based on the comparison of the calculation results with the set of precise experimental data that we have made available in the literature we conclude that we can predict free energy differences with a precision comparable to that of experiments, provided the mechanism of association and structural relaxations are taken into account.
Dynamics of protein-ligand interactions (P.-L. Chau)
An 1-ns unbinding trajectory of retinol from the bovine serum retinol-binding protein has been obtained from molecular dynamics simulations. The behaviour of water during ligand unbinding has never been studied in detail. In this study, a new method for defining a binding site was developed, the water molecules involved in the binding site were located, and their movements examined during unbinding. It was found that there were only small changes in the binding site. The number of water molecules inside the binding site decreased, and there were re-arrangements of the hydrogen bonds, during unbinding. This work represents the first detailed study of the behaviour of water during an unbinding process.
The dynamics of protein-protein complex formation (R. Grünberg, J. Leckner, M. Nilges)
One of the major problems when predicting protein-protein complexes is posed by structural transitions between the free and bound conformations. Protein-protein docking algorithms need to rely on various simplifications, since fully flexible docking is computationally not feasible. To improve docking procedures, detailed knowledge about dynamic properties of binding surfaces is necessary. While static properties like shape complementarity and differences between free and bound structures have been studied in quite some detail, little is known about how binding affects - and is affected by - the dynamics of the protein partners. We selected a set of 17 protein-protein complexes for which the three-dimensional structures of both free components and the complex are available. For each of these 51 systems we performed molecular dynamics simulations in explicit water. Surprisingly, in many cases binding does not restrict the motion of interface residues or of the proteins as a whole. In the majority of cases, average fluctuations of the bound components are similar or higher when compared to the free trajectories. Such increased flexibilities have also been observed experimentally in complexes of Calmodulin, protein L9, Protein kinase A, and others. Moreover, we often observe that already the free receptor and ligand interfaces show complementary dynamic behaviour.
Interaction between 5-HT and the 5-HT3 receptor (P.L. Chau, N. Duclert-Savatier, M. Nilges; collaboration with Sarah Lummis (Department of Biochemistry, University of Cambridge))
This study is a continuation of previous work on the predicted complex of 5-HT and a homology model of the 5-HT3 receptor, which showed good agreement with experimental findings (D.C. Reeves, M.F.R. Sayed, P.-L. Chau, K.L. Price and S.C.R. Lummis (2003) Biophysical Journal 84 2338--2344). The properties of the modelled complex were further studied by molecular dynamics calculations. A water sphere of 12 Å was placed around the binding site, and the dynamic properties of 5-HT studied using Langevin dynamics. The dynamics will permit a detailed analysis of the strength of the different interactions between the ligand and the protein, which will be compared to the experimental results obtained in the laboratory of Sarah Lummis.
Hydrophobic hydration (P.-L. Chau)
The influence of curvature on hydrophobic hydration was studied by molecular dynamics calculations. Three hemispherical objects of radius 6.5 Å, 9.3 Å and 12.2 Å were constructed from tessellated icosahedra where the vertices were hydrophobic "atoms". These objects were hydrated, and molecular dynamics simulations were performed on the solvated system. Analysis of the trajectories demonstrated that water molecules around convex solutes possess different configurations and exhibit different translational and rotational dynamics from water molecules in the bulk region. No qualitative difference was observed in the alteration of these structural and dynamic properties as the hemispheres became larger.
Interpretation of experimental NMR data
A new principle for NMR structure determination (W. Rieping, M. Habeck, M. Nilges)
The principal determinants of uncertainties in NMR structures are imperfections in the experimental data and inaccuracies of theoretical models. We are developing a new structure determination principle, called Inferential Structure Determination (ISD), based on Bayesian probability theory, in order to deal with uncertainties of experimental data and theoretical models in a fundamental way. The approach is very general and not limited to NMR structure determination. One key advantage is that all unknown quantities, not only atomic coordinates, can be estimated from the data. Accordingly, ISD contains no free parameters, thus reducing human bias to a minimum: calculated structures depend only on the experimental data and additional prior knowledge about physical interactions of the system, implemented in a molecular dynamics force field. The estimated precision of the structure is only determined by the data, as it ideally should be. Using this approach, we can for the first time derive statistically meaningful figures of merit for NMR structures (atomic uncertainties). The Bayesian analysis also permits us to evaluate the consistency of the data, using predictive distributions. In addition, we are working on improvements of the data model used to directly include parameters describing internal dynamics into the calculation.
Larges-scale structural models and genomic analyses (E. Yeramian)
In the prolongation of previous work, the developments (based on collaborations) are following four complementary directions: 1) Algorithmic methodologies, for the implementation of realistic large-scale structural models. Usually such implementations are limited by computational complexities relative to the long-range effects. The new extensions concern the problems of RNA secondary structures and sequence alignments with non-linear gap penalties (with E. Debonneuil); 2) Elaboration of physical bases for large-scale structural models. The recent developments concern notably supercoiled DNA (with H. Orland and T. Garel); 3) Genomic analyses and gene identifications on structural bases. Such analyses (with L. Jones) being implemented now in a new web-site 'pbga.pasteur.fr' (PBGA: Physics-Based-Genomic-Analyses) and 4) Genome comparisons for the evolutionary description of phylogenies (with F. Tekaia).
Keywords: protein structure prediction, protein interactions, docking, molecular modelling, genome analysis