Loop Refinement by Localized Sampling and Minimization (SLAM) Vageli Coutsias, Matt Jacobson, Michael Wester, Lan Hua.

Slides:



Advertisements
Similar presentations
Protein Structure C483 Spring 2013.
Advertisements

Review of Basic Principles of Chemistry, Amino Acids and Proteins Brian Kuhlman: The material presented here is available on the.
Rosetta Energy Function Glenn Butterfoss. Rosetta Energy Function Major Classes: 1. Low resolution: Reduced atom representation Simple energy function.
Protein Structure – Part-2 Pauling Rules The bond lengths and bond angles should be distorted as little as possible. No two atoms should approach one another.
The amino acids in their natural habitat. Topics: Hydrogen bonds Secondary Structure Alpha helix Beta strands & beta sheets Turns Loop Tertiary & Quarternary.
Protein Threading Zhanggroup Overview Background protein structure protein folding and designability Protein threading Current limitations.
Chemical Biology 03 BLOOD
Protein-a chemical view A chain of amino acids folded in 3D Picture from on-line biology bookon-line biology book Peptide Protein backbone N / C terminal.
1 Levels of Protein Structure Primary to Quaternary Structure.
Ensemble Results of PIM1 PIM PIM Ensemble Results of GSK3 GSK GSK GSK
Summary Protein design seeks to find amino acid sequences which stably fold into specific 3-D structures. Modeling the inherent flexibility of the protein.
Protein Basics Protein function Protein structure –Primary Amino acids Linkage Protein conformation framework –Dihedral angles –Ramachandran plots Sequence.
A Kinematic View of Loop Closure EVANGELOS A. COUTSIAS, CHAOK SEOK, MATTHEW P. JACOBSON, KEN A. DILL Presented by Keren Lasker.
Amino Acids C483 Spring Amino Acid Structure Alpha carbon Sidechain Proteins peptides.
A PEPTIDE BOND PEPTIDE BOND Polypeptides are polymers of amino acid residues linked by peptide group Peptide group is planar in nature which limits.
Proteins: Levels of Protein Structure Conformation of Peptide Group
Proteins account for more than 50% of the dry mass of most cells
Macromolecular structure
Practical session 2b Introduction to 3D Modelling and threading 9:30am-10:00am 3D modeling and threading 10:00am-10:30am Analysis of mutations in MYH6.
What are proteins? Proteins are important; e.g. for catalyzing and regulating biochemical reactions, transporting molecules, … Linear polymer chain composed.
COMPARATIVE or HOMOLOGY MODELING
Protein Secondary Structure Lecture 2/19/2003. Three Dimensional Protein Structures Confirmation: Spatial arrangement of atoms that depend on bonds and.
Proteins: Secondary Structure Alpha Helix
Proteins. Proteins? What is its How does it How is its How does it How is it Where is it What are its.
©CMBI 2006 Amino Acids “ When you understand the amino acids, you understand everything ”
Protein “folding” occurs due to the intrinsic chemical/physical properties of the 1° structure “Unstructured” “Disordered” “Denatured” “Unfolded” “Structured”
RNA Secondary Structure Prediction Spring Objectives  Can we predict the structure of an RNA?  Can we predict the structure of a protein?
ProteinShop: A Tool for Protein Structure Prediction and Modeling Silvia Crivelli Computational Research Division Lawrence Berkeley National Laboratory.
Bioinformatics: Practical Application of Simulation and Data Mining Protein Folding I Prof. Corey O’Hern Department of Mechanical Engineering & Materials.
Protein Structure Stryer Short Course Chapter 4. Peptide bonds Amide bond Primary structure N- and C-terminus Condensation and hydrolysis.
STRUCTURE CALCULATIONS OF PROTEIN SURFACE SEGMENTS: MONTE CARLO SIMULATED ANNEALING WITH SCALED COLLECTIVE VARIABLES AND FORCE CONSTANT ANNEALING Sergio.
Department of Mechanical Engineering
Lecture 1: Fundamentals of Protein Structure
Chap. 4. Problem 1. Part (a). Double and triple bonds are shorter and stronger than single bonds. Because the length of a peptide bond more closely resembles.
Structure prediction: Homology modeling
Protein Design with Backbone Optimization Brian Kuhlman University of North Carolina at Chapel Hill.
Structure prediction: Ab-initio Lecture 9 Structural Bioinformatics Dr. Avraham Samson Let’s think!
Amino Acids ©CMBI 2001 “ When you understand the amino acids, you understand everything ”
Protein Structure and Bioinformatics. Chapter 2 What is protein structure? What are proteins made of? What forces determines protein structure? What is.
Protein backbone Biochemical view:
CS-ROSETTA Yang Shen et al. Presented by Jonathan Jou.
Structural organization of proteins
Protein Structure BL
Visualization Homework – Take II
The heroic times of crystallography
Structure of the Rho Family GTP-Binding Protein Cdc42 in Complex with the Multifunctional Regulator RhoGDI  Gregory R. Hoffman, Nicolas Nassar, Richard.
Beta sheets come in two flavors: parallel (shown on this slide) and anti parallel. The geometry of the individual beta strandis are almost identical in.
Hierarchical Structure of Proteins
Lecture 5 Protein Structure.
Enzyme Kinetics & Protein Folding 9/7/2004
Wenqing Xu, Amish Doshi, Ming Lei, Michael J Eck, Stephen C Harrison 
-Primary and Secondary Structure-
Volume 86, Issue 6, Pages (June 2004)
Rosetta: De Novo determination of protein structure
Levels of Protein Structure
Protein structure prediction.
Near-Atomic Resolution for One State of F-Actin
Structural and Dynamic Properties of the Human Prion Protein
Volume 8, Issue 4, Pages (April 2001)
Complementarity of Structure Ensembles in Protein-Protein Binding
Volume 15, Issue 6, Pages (December 2001)
Structure of the Rho Family GTP-Binding Protein Cdc42 in Complex with the Multifunctional Regulator RhoGDI  Gregory R. Hoffman, Nicolas Nassar, Richard.
Feng Ding, Sergey V. Buldyrev, Nikolay V. Dokholyan 
Volume 86, Issue 6, Pages (June 2004)
Protein structure prediction
Structure of an IκBα/NF-κB Complex
Insights from Free-Energy Calculations: Protein Conformational Equilibrium, Driving Forces, and Ligand-Binding Modes  Yu-ming M. Huang, Wei Chen, Michael J.
Hydrophobic Core Formation and Dehydration in Protein Folding Studied by Generalized-Ensemble Simulations  Takao Yoda, Yuji Sugita, Yuko Okamoto  Biophysical.
Tertiary structure of an immunoglobulin-like domain from the giant muscle protein titin: a new member of the I set  Mark Pfuhl, Annalisa Pastore  Structure 
Morgan Huse, Ye-Guang Chen, Joan Massagué, John Kuriyan  Cell 
Presentation transcript:

Loop Refinement by Localized Sampling and Minimization (SLAM) Vageli Coutsias, Matt Jacobson, Michael Wester, Lan Hua

(1) loops to be refined - given or by comparison with homologous proteins [ for homology modeling only: use PRIME to get templates. Identify loops that show most variability among different templates. ] (2) use SPLAT to sample around the loops identified in (1). Sampling is restricted to Ramachandran allowed regions (these are sampled with probability-usually.001 threshold) and screened for sterics with variable steric cutoff. Too large screens exclude interesting contacts from forming. Too small give high clash energies. (3) use PLOP to minimize energy of the loop regions (including residues within a certain cutoff distance, mostly set at 7A). (4) select lowest energy candidates without any obvious flaws (i.e. loops away in solvent, obvious burials of polar residues or hydrophobics blatantly pointing outwards) and resample around them with windowed sampling - usually +/-30 degrees, except for certain residues I wanted to maintain. I tried different strategies for the windowed sampling. (5) iterate steps 3-4. Typically looked for a funnel-like distribution of RMSD from a reasonable minimum vs. energy to decide convergence. Only used systematically (but not completely due to time pressure) for R488.

Results R432: rmsd to native 1.0 (vs. 1.3 template) R453: higher rmsd / correct secondary R488: higher rmsd T0479: rmsd to native 1.3

Conclusion Although sampling seems to have revealed the shape of the loop including occasional secondary structure elements, special contacts & salt bridges formed that minimized well but forced the sampling into ranges far from the native. In subsequent work we should: (1)include a stage of localized move MC to properly weigh minimum states by free energy estimates. Various constraints will allow to focus sampling and raise efficiency. (2) Must include a Jacobian-based pruning of data to increase efficiency by cutting down on minimizations. (3) Explore other global search & optimization methods to produce a map of the low minima. (4) Add enforcing of position and orientation constraints

R432 = 3dai Refinement of loop Small helix end perturbation Broad sampling of loop Cyan (model 1) is at 1.0A to native (purple) Template (white) is at 1.3A Loop is at 2.4 A (cyan) and 3.8 A (white) Strategy: lowest energy structures from 1 st batch start own narrower sampling branches. Best results of second iteration sampled again. Now low energy extension of helix at 90 was found, and it was selected in some of the new seed runs. Lowest energy results from the fourth stage were chosen. R432

R453 = 3ded Loop in space. Found structure with a single-turn helix. Appears in chain A Template, although closer in RMSD, lacks this structure. W:native, P:template, C:best model Fig. 1: matched to chain D Fig. 2: matched to chain A R453

Small twist around res35 (which is a Proline and res 37 is also a Proline) in both native and model structures. Interestingly, in the model structures the carbonyl oxygen of Pro37 seems to form hydrogen bond with OH of Tyr53, however no such hydrogen bond is formed in native structure and the initial best model. Probably such hydrogen bond formation pulls the loop turn (res35-40) in the model structure away from native structure. R453

R488 with ligand (D). 6k set, full 360 sampling of all residues (11-19) of TR488 with 6-ALA ligand (E)3k set, 360 at pivots (1-5-7), +/-30 at LEU18, GLY19 (to preserve orientation of LEU sidechain), +/-60 at other residues (F) 6k set, 360 at pivots (same), +/-30 at all other residues Structure of minimum stable; a few structures with slightly lower energy ~2kcal, but ruined secondary structures (helices or strands un-made) TOP candidate: T488_f_04930-opt.pdb Shown with its precursors, d and e Together with other f-batch structures of similar shape, but slightly higher energies R488

d-, e-1016, f-4930 and other low-energy F-structures; well converged shape R488

Although this model represents the height of the preformance of the algorithm, it had a twist, which made the effort totally speculative: as a PDZ domain, it ought to have a binding pocket. My first set of efforts did not take that into account, and I simply sampled/minimized quite extensively, iterating several times until I arrived at a very robust-seeming loop. Then we noticed that the binding pocket was too tight, so we introduced a putative ligand and sampled again to allow for its presense. I also restricted sampling to conformations that had the LEU18 pointing inward. The sidechain of LEU18 is pointing inward in native structure, initial best model and model structures. Asp13 points inward in both refined models, probably forming salt bridges with Lys11 (the distance between OD of Asp13 and NZ of Lys 11 is around 2.7 angstrom), which twist the loop inward. However Asp13 points outward in native structure and initial best model, and does not form salt bridge with Lys11. Native (W), model 1 (C), model 2 (P), template (Y)

TO479 = 3dkz Model T0479_3 (model 3) has the smallest rmsd (1.3 A) and T0479_4 (mod 4) the largest (2.3 A). Refining three loops: used various combinations of minima from each refinement, on the assumption that the loops did not interact. Homology modeling using PRIME. Subsequent loop refinement using SPLAT(sampling)/PLOP(minimization). (fig1) Native (W) vs models 3 (Y), 4 (R) (fig2) Native (W), 1(P), 2(C), 3(Y), 4(R), 5(B) T0479