From Protein Structure to Function with Bioinformatics.pdf

3 Comparative Protein Structure Modelling 75such as the Replica Exchange Monte Carlo simulation can be applied (Han et al.2008; Qian et al. 2004).3.2.4.4 Modelling of Proteins and Complexes with Additional,Experimental RestraintsSome comparative modelling techniques are able to incorporate constraints orrestraints derived from a number of different sources other than the homologous templatestructure. For example, restraints could be provided by rules for secondarystructure packing (Cohen et al. 1989), analyses of hydrophobicity (Aszodi and Taylor1994) and correlated mutations (Taylor and Hatrick 1994), empirical potentials ofmean force (Sippl 1995), nuclear magnetic resonance experiments (Sutcliffe et al.1992), or from experiments on chemical cross-linking, spin and photoaffinity labelling(Orr et al. 1998), hydrogen/deuterium exchange coupled with mass spectrometry(Xiao et al. 2006), hydroxyl radical footprinting (Kiselar et al. 2003), fluorescencespectroscopy, image reconstruction in electron microscopy (Topf et al. 2008), sitedirectedmutagenesis (Boissel et al. 1993) etc. In this way, a comparative model,especially in the difficult cases, could be improved by making it consistent with availableexperimental data and with more general knowledge about protein structure.In the past, comparative modelling relied mostly on template information and statistically-derivedrestraints from known protein structures and sequences. But it isexpected that with the advances of large scale genetic and proteomics techniquesmore and more experimentally derived restraints will be available for automaticincorporation in the modelling process. In addition to delivering more accurate models,this should particularly facilitate the modelling of protein complexes andassemblies.A systematic approach to tackle the modelling of large protein complexes withthe aid of experimental restraints was developed for the modelling of the nuclearpore complex, the largest known protein complex in the cell that consist of 456proteins (Alber et al. 2008). The approach integrated a wealth of experimentalinformation. For instance, quantitative immunoblotting determined the stoichiometry,while hydrodynamics experiments provided insight about the approximateshape and excluded volume of each nucleoporins; immuno–EM helped incoarse localization of nucleoporins; affinity purification determined the compositionof complexes; cryo-EM and bioinformatics analysis uncovered locations oftransmembrane segments and overlay experiments gave information on directbinary interactions. All these data inputs were integrated in a hierarchical processthat combined comparative modelling, threading, rigid and flexible dockingtechniques. The ultimate goal of the data integration is to convert all availableexperimental information into spatial restraints that can guide the generalizedmodelling procedure. The procedure is flexible to combine entities of variousrepresentations and resolutions (for instance atoms, atomistic models of proteins,symmetry units or whole assemblies) and optimization procedures (Alber et al.2007a, b, 2008). This and similar efforts will leverage benefits simultaneously