340 likes | 448 Vues
Explore the information encoded within the surface of a protein-protein complex to predict hot spots. Utilize computational methods for biochemically relevant results and visualize the interface hierarchy. Further work includes refinement of predictions and investigation into protected regions. Acknowledge contributions from experts in the field.
E N D
Information Within the Interface Surface of a Protein-Protein Complex Yih-En Andrew Ban Duke University Biochemistry & Computer Science
Problem Statement • Why & how do proteins dock?
Objectives • Construct a representation: • Aids in biochemical analysis. • Supplants the need for experiment. • Issues: • Time • Accuracy
Typical Methods • Empirical Force Field • Molecular dynamics • Monte Carlo simulation • Simplification/Hierarchy • Substructure manipulation • Rotamer libraries
Biophysical Models Energy Geometry Biochemical Meaning
Relevance • Establish a readily useable biochemical result. • Hot-spot prediction.
Intuition • Medial surface captures the essentials of the interaction. • Regions of importance are protected in some way.
Concepts • Voronoi diagram • Delaunay triangulation • Alphashapes • Topological Persistence
In Practice • Construct Delaunay triangulation • Construct Alphashape filtration • Orders simplices based upon size • Apply pairing algorithm on the Alphashape filtration • Identification of protected regions • Construct retraction hierarchy • Removal of initial unprotected region • Removal of protected regions • Construction of interfaces
Seal Function • where s is the size of the orthogonal ball of the triangle • where u is the size of the orthogonal ball of the tetrahedra
Nomenclature • Gate = seal triangle • Flood = set of triangles and tetrahedra that are deleted and retracted • Trench = trivial collapse
Hot-Spot Function • where R is a residue • p0 .. pkare the polygons of R • S is the interface surface
Prediction • Kortemme & Baker (2002) • 19 protein-protein complexes • 234 residues • 71 hot, 163 neutral • Interface surface generation • Heavy atoms only • h(R) theshold = 3.75 • ddG threshold = 2.0 kcal/mol
Competing Method • Kortemme & Baker (2002) • virtual alanine scanning • simple force-field model • rotamer library • Monte Carlo Optimization • full atomic detail • ddG threshold = 1.0 kcal/mol
Conclusion • Interface surface • Biochemically relevant • Reasonable model for analyzing protein-protein interactions • Information encoded within the interface is substantial – hot spots can be predicted!? • Further work • Refinement of h(R) • Investigation into protected regions • Visualization • etc…
Acknowledgements • Herbert Edelsbrunner • Johannes Rudolph