From Protein Structure to Function with Bioinformatics

Table of Contents Part 1: Generating and inferring structures 1 Ab initio protein structure prediction 1.1 Introduction 1.2 Energy functions 1.2.1 Physics-based energy functions 1.2.2 Knowledge-based energy function combined with fragments 1.3 Conformational search methods 1.3.1 Monte Carlo simulations 1.3.2 Molecular dynamics 1.3.3 Genetic algorithm 1.3.4 Mathematical optimization 1.4 Model selection 1.4.1 Physics-based energy function 1.4.2 Knowledge-based energy function 1.4.3 Sequence-structure compatibility function 1.4.4 Clustering of decoy structure 1.5 Remarks and discussion 2 Fold Recognition 2.1 Introduction 2.1.1 The importance of blind trials: the CASP competition 2.1.2 Ab initio structure prediction versus homology modelling 2.1.3 The limits of fold space 2.1.4 A note on terminology: ???threading??? and ???fold recognition??? 2.2 Threading 2.2.1 Knowledge-based potentials 2.2.2 Finding an alignment 2.2.3 Heuristics for alignment 2.3 Remote homology detection without threading 2.3.1 Using predicted structural features 2.3.2 Sequence profiles and hidden Markov models 2.3.3 Fold Classification and Support Vector Machines 2.3.4 Consensus approaches 2.3.5 Traversing the homology network 2.4 Alignment accuracy, model quality and statistical significance 2.4.1 Algorithms for alignment generation and assessment 2.4.2 Estimation of statistical significance 2.5 Tools for fold recognition on the web 2.6 The future 3 Comparative protein structure modelling 3.1 Introduction 3.1.1 Structure determines function 3.1.2 Sequences, structures, structural genomics 3.1.3 Approaches to protein structure prediction 3.2 Steps in comparative protein structure modelling 3.2.1 Searching for structures related to the target sequence 3.2.2 Selecting templates 3.2.3 Sequence to structure alignment 3.2.4 Model building 3.2.5 Model evaluation 3.3 Performance of comparative modelling 3.3.1 Accuracy of methods 3.3.2 Errors in comparative models 3.4 Applications of comparative modelling 3.4.1 Modelling of individual proteins 3.4.2 Comparative modelling and the Protein Structure Initiative 3.5 Summary 4 Membrane protein structure prediction 4.1 Introduction 4.2 Structural classes 4.2.1 Alpha-helical bundles 4.2.2 Beta-barrels 4.3 Membrane proteins are difficult to crystallise 4.4 Databases 4.5 Multiple sequence alignments 4.6 Transmembrane protein topology prediction 4.6.1 Alpha-helical proteins 4.6.2 Beta-barrel proteins 4.6.3 Whole genome analysis 4.6.4 Data sets, homology, accuracy and cross-validation 4.7 3D structure prediction 4.8 Future developments 5 Bioinformatics approaches to the structure and function of intrinsically disordered proteins 5.1 The concept of protein disorder 5.2 Sequence features of IDPs 5.2.1 The unusual amino acid composition of IDPs 5.2.2 Sequence patterns of IDPs 5.2.3 Low sequence complexity and disorder 5.3 Prediction of disorder 5.3.1 Prediction of low-complexity regions 5.3.2 Charge-hydropathy plot 5.3.3 Propensity-based predictors 5.3.4 Predictors based on the lack of secondary structure 5.3.5 Machine learning algorithms 5.3.6 Prediction based on contact potentials 5.3.7 A reduced alphabet suffices to predict disorder 5.3.8 Comparison of disorder prediction methods 5.4 Functional classification of IDPs 5.4.1 Gene Ontology-based functional classification of IDPs 5.4.2 Classification of IDPs based on their mechanism of action 5.4.3 Function-related structural elements in IDPs 5.5 Prediction of the function of IDPs 5.5.1 Correlation of disorder pattern and function 5.5.2 Predicting short recognition motifs in IDRs 5.5.3 Prediction of MoRFs 5.5.4 Combination of information on sequence and disorder: phosphorylation sites and CaM binding motifs 5.5.5 Flavours of disorder 5.6 Limitations to IDP function prediction 5.6.1 Rapid evolution of IDPs 5.6.2 Sequence independence of function and fuzziness 5.6.3 Good news: conservation and disorder 5.7 Conclusions Part 2: From structures to functions 6 Function diversity within folds and superfamilies 6.1 Defining function 6.2 From fold to function 6.2.1 Definition of a fold 6.2.2 Prediction of function using fold relationships 6.3 Function diversity between homologous proteins 6.3.1 Definitions 6.3.2 Evolution of protein superfamilies 6.3.3 Function divergence during protein evolution 6.4 Conclusion 7 Predicting protein function from surface properties 7.1 Surface descriptors 7.1.1 The van der Waals surface 7.1.2 Molecular surface (solvent excluded surface) 7.1.3 The solvent accessible surface 7.2 Surface properties 7.2.1 Hydrophobicity 7.2.2 Electrostatics properties 7.2.3 Surface conservation 7.3 Function predictions using surface properties 7.3.1 Hydrophobic surface 7.3.2 Electrostatic surface 7.3.3 Surface conservation 7.3.4 Combining surface properties for function prediction 7.4 Protein-ligand interactions 7.4.1 Properties of protein-ligand interactions 7.4.2 Predicting binding site locations 7.4.3 Predictions of druggability 7.4.4 Annotation of ligand binding sites 7.5 Protein-protein interfaces 7.5.1 Properties of protein-protein interfaces 7.5.2 Hot-spot regions in protein interfaces 7.5.3 Predictions of interface location 7.6 Summary 8 3D Motifs 8.1 Background and significance 8.1.1 What is function? 8.1.2 3D Motifs: Definition and scope 8.2 Overview of Methods 8.2.1 Motif discovery 8.2.2 Motif description and matching 8.2.3 Interpretation of results 8.3 Specific Methods 8.3.1 User-defined motifs 8.3.2 Motif discovery 8.4 Related methods 8.4.1 Hybrid (point-surface) descriptions 8.4.2 Single-point-centred descriptions 8.5 Docking for functional annotation 8.6 Discussion 8.7 Conclusions 9 Protein dynamics: from structure to function 9.1 Molecular dynamics simulations 9.1.1 Principles and approximations 9.1.2 Applications 9.1.3 Limitations 9.2 Principal component analysis 9.3 Collective coordinate sampling algorithms 9.3.1 Essential dynamics 9.3.2 TEE-REX 9.4 Methods for functional mode prediction 9.4.1 Normal mode analysis 9.4.2 Elastic network models 9.4.3 CONCOORD 9.5 Summary and outlook 10 Integrated servers for structure-informed function prediction 10.1 Introduction 10.1.1 The problem of predicting function from structure 10.1.2 Structure-function prediction methods 10.2 ProKnow 10.2.1 Fold matching 10.2.2 3D motifs 10.2.3 Sequence homology 10.2.4 Sequence motifs 10.2.5 Protein interactions 10.2.6 Combining the predictions 10.2.7 Prediction success 10.3 ProFunc 10.3.1 ProFunc???s structure-based methods 10.3.2 Assessment of the structural methods 10.4 Conclusion 11. Case studies: Function predictions of structural genomics results 11.1 Introduction 11.2 Large scale function prediction case studies 11.3 Some specific examples 11.4 Community annotation 11.5 Conclusions 12. Prediction of protein function from theoretical models 12.1 Background 12.2 Protein models as a community resource 12.2.1 Model quality 12.2.2 Databases of models 12.3 Accuracy and added value of model-derived properties 12.3.1 Implementation 12.4 Practical application 12.4.1 Plasticity of catalytic site residues 12.4.2 Mutation mapping 12.4.3 Protein complexes 12.4.4 Function predictions from template-free models 12.4.5 Prediction of ligand specificity 12.4.6 Structure modelling of alternatively spliced isoforms 12.4.7 From broad function to molecular details 12.5 What next? Index