DragonStar-5.ppt

Computational Methods for Structural Bioinforamtics and Computational Biology (5) (Empirical Potential Function for Structure Prediction  and for Protein Design) Today’s Lecture Theoretical framework for developing knowledge based potential Statistical potential function: Two-body, three-body and cooperativity Reference state, null model and motif finding Optimized potential function: Global nonlinear fitness landscape Anfinsen’s Experiments In vitro refolding studies Christian Anfinsen: ribonuclease can refold spontaneously to active form Same for other proteins Sequence contains all information for folding to its native form Anfinsen’s Hypothesis Thermodynamic hypothesis of protein folding: A native protein folds into a three-dimensional structure in equilibrium, The state of the whole protein-solvent system corresponds to the global minimum of free energy under physiological conditions. Potential function Protein folding. Divide-and-conquer: Decoy generation. Decoy discrimination. Guide conformation generation Keasar and Levitt, 03; Bonneau Baker, 02; Skolnick et al, 97. Protein design. Novel or enhanced fucntion. Novel fold: Top7 by Baker group. Key Questions: When does one know a good solution is found? How to generate solutions that are good? How to search the conformational and sequence space? Need scoring function! (potential function, energy function) Physical based potential Physical based potential: force-filed, semi-empirical Parameters: matching quantum mechanical calculations to measured thermodynamic properties of small molecules . Generalized to proteins Widely used in folding simulations (Lazaridis and Karplus, 2000; Duan and Kollman, 1998) Difficulty: Requires full atomic model, expensive to compute. Physical model may not fully capture all othat is important.. Knowledge based potential An alternative approach: Extracting parameters from database of protein structure Only implicitly incorporates physical forces: