Computational prediction of protein-protein interactions Rong Liu
Quaternary structure
Types of protein-protein interactions Homo-oligomers vs. hetero-oligomers Permanent vs. transient interactions Strong transient Weak transient Covalent vs. non-covalent interactions Classification based on function enzyme-inhibitor antibody-antigen Others (e.g. hormone-receptor, signaling-effector)
Experimental methods to detect PPI Yeast two hybrid (Y2H) Tandem affinity purification coupled to mass spectrometry (TAP-MS) Co-inmunoprecipitation (CoIP) Protein microarrays Phage display Surface plasmon resonance ……
Reliability of high-throughput methods
PPI database
A sample of DIP protein table
List of interacting partners
Graphic representation of interactions Nodes are proteins Edges are PPIs The center node is DIP:1143N Edge width encodes the number of independent experiments identifying the interactions. Green (red) is used to draw core (unverified) interactions. Click on each node (edge) to know more about the protein (interaction).
Techniques to study the protein complex structures X-ray crystallography Nuclear magnetic resonance spectroscopy Electron Microscopy
Header of PDB file
Format of PDB file
Preparation of PPI and non-PPI datasets PPI dataset (Gold standard dataset) Data from multiple database At least two separate publications Each of these publications needs to have a binary evidence code Non-PPI dataset Random selection from all possible protein pairs Proteins come from different sub-localization
The first non-PPI database
PPI prediction based on homology
InParanoid8
Genome context-based methods
Domain association-based method
Domain combination
Machine learning-based method
Feature representation of amino acid sequences
Protein feature server
Validation of the predicted PPIs
Protein-protein binding interface
Hotspots in binding interface (ΔΔG >2kcal/mol)
Definition of binding interface Define surface residue (DSSP, NACCESS) Define interface residue Distance-based method Solvent accessible surface area-based method
Format of DSSP file
Characteristics analysis of binding interface
Features of transient and obligate interactions
Features used to predict PPI binding interface Sequence conservation Propensity of residue types in binding regions Secondary structure Solvent accessibility Protrusion index Side-chain conformational entropy
Position specific scoring matrix and neighborhood
Training and testing Cross-validation and independent test Balanced positive and negative samples Evaluation measures
State-of-the-arts of feature-based prediction
Similarity between binding interfaces
Protein interface conservation across structure space
Performance comparison between different algorithms
Hybrid method
Residue interaction network
Network-based features Degree centrality Closeness centrality Betweenness centrality Clustering coefficient
Protein complexes and small-world networks
Network-based features of other binding sites
Graph-based interface alignment
InterPreTS: protein Interaction Prediction through Tertiary Structure
Structure-based prediction of protein–protein interactions on a genome-wide scale
Protein Docking
Procedure of protein docking
Search of conformations
Scoring function
Docking programs and benchmark
Evaluation measures
Structure visualization tools FeatureRasMolCn3DPyMolSWISS- PDBViewer Chimera ArchitectureStand-AlonePlug-inStand-AloneWeb-enabled Manipulation Power LowHigh Hardware Requirements Low/ModerateHigh ModerateHigh Ease of UseHigh; command line Moderate HighModerate;GUI +command line Special FeaturesSmall Size; easy install Powerful GUI GUI; ray tracing Powerful GUIGUI; collaboration Output QualityModerateVery highHigh Very high DocumentationGood LimitedGoodVery good SupportOnline; Users groups SpeedHighModerate Moderate/Slow OpenGL SupportYes
Pymol
Application of PPI network