A Rational Approach to Receptor-Flexible Docking: Method and Validation

Transcription

1 For audio: Dial-in Information: Toll Free +1 (866) Toll +1 (770) Participant Code: ctober 2007 A Rational Approach to Receptor-Flexible Docking: Method and Validation C. M. (Venkat) Venkatachalam, Ph. D. Fellow, Accelrys

2 verview Receptor Changes in Ligand Binding Flexible Docking Discovery Studio 2.0 Architecture of DS and PipeLine Pilot Key Pipeline Pilot Components used in Flexible Docking ChiFlex, CatConf, LibDock, ChiRotor, CDCKER Flexible Docking Workflow Results of Cross Docking 2

3 Hydrogen bonding interactions in Thymidine Kinase within 4 angstroms of the ligand Hydrogen bonding network Ligand Gln125 (2 hbs) Ligand Glu225 Glu83 Arg222 Ile97 Tyr101 Gln125 Ala168 Arg163 Tyr172 Arg222 Glu kim complex from xray

4 Variation in Side chains in Estrogen: 1err vs 3ert chi = chi(3ert) -chi(1err) residue chi1 chi2 LEU LEU THR ARG352 LEU354 GLU err ILE TRP ARG MET HIS LEU LEU TYR537 ASP ert LEU

5 Flexible Docking: Needs and Challenges Scientific Need: Increased accuracy of docked poses Challenges: Ignoring receptor flexibility during docking leads to inaccurate poses The problem is magnified in vhts There is often a complex network of interactions between a protein and a bound ligand over many residues In existing docking methods, receptor conformational search needs to be repeated for each small molecule during vhts For real vhts applications, flexible docking needs to be fully automated Complex network of interactions in Thymidine Kinase (PDB ID 1kim) 5

6 Discovery Studio and Pipeline Pilot 6 The Flexible Docking protocol requires a Pipeline Pilot Client Can be run in Pipeline Pilot or Discovery Studio 2.0

7 Key Components used to build a Flexible Docking Workflow ChiRotor Side chain reconstruction ChiFlex Generates low energy side conformations CatConf Ligand Conformation generation LibDock Docks a ligand conformation rigidly to receptor site CDCKER Ligand refinement in the presence of the receptor ChiRotor/ChiFlex V. Z. Spassov, L. Yan, P. K. Flook, The Dominant Role of Side-chain Backbone Interactions in Structural Realization of Amino-acid Code. ChiRotor: a Side-chain Prediction Algorithm Based on Side-chain Backbone Interactions, Protein Science 16, 1-13 (2007). CDCKER G. Wu, D. H. Robertson, C. L. Brooks III and M. Veith, Detailed Analysis of Grid-Based Molecular Docking: A Case Study of CDCKER- A CHARMm-Based MD Docking Algorithm, J. Comp. Chem. 24, , J. A. Erickson, M. Jalaie, D. H. Robertson, R. A. Lewis and M. Vieth, Lessons in Molecular Recognition: The Effects of Ligand and Protein Flexibility on Molecular Docking Accuracy, J. Med. Chem., 47 (1), 45-55, 2004 LibDock D. Diller and K. Mertz, PRTEINS: Structure, Function, and Genetics 43, (2001) D. Diller and Li, R. J. Med. Chem. 46, (2003) 7

8 Key Components used to build a Flexible Docking Workflow ChiRotor Side chain reconstruction ChiFlex Generates low energy side conformations CatConf Ligand Conformation generation LibDock Docks a ligand conformation rigidly to receptor site CDCKER Ligand refinement in the presence of the recptor ChiRotor/ChiFlex V. Z. Spassov, L. Yan, P. K. Flook, The Dominant Role of Side-chain Backbone Interactions in Structural Realization of Amino-acid Code. ChiRotor: a Side-chain Prediction Algorithm Based on Side-chain Backbone Interactions, Protein Science 16, 1-13 (2007). CDCKER G. Wu, D. H. Robertson, C. L. Brooks III and M. Veith, Detailed Analysis of Grid-Based Molecular Docking: A Case Study of CDCKER- A CHARMm-Based MD Docking Algorithm, J. Comp. Chem. 24, , J. A. Erickson, M. Jalaie, D. H. Robertson, R. A. Lewis and M. Vieth, Lessons in Molecular Recognition: The Effects of Ligand and Protein Flexibility on Molecular Docking Accuracy, J. Med. Chem., 47 (1), 45-55, 2004 LibDock D. Diller and K. Mertz, PRTEINS: Structure, Function, and Genetics 43, (2001) D. Diller and Li, R. J. Med. Chem. 46, (2003) 8

9 ChiRotor FlowChart Protein 3D Structure Select a Set of n Residues For Refinement Remove all side chain atoms of selected residues Start loop for i from 1 to n Choose Residue i Sample side chain conformations of residue i varying χ 1 Energy Minimize side chain atoms of residue i in CHARMm Save 2 Best Conformations for residue i End loop for i Construct complete structure using lowest energy conformer of each residue. Energy minimize all selected side chains Start loop for i from 1 to n Replace side chain conformation i with the 2 nd conformer and energy minimize Accept the structure if energy is lower. End loop for i 9 utput: 2n partial structures utput: 1 Lowest Energy structure

10 Side Chain Conformations in Thymidine Kinase 1kim 1qhi 1ki4 1e2m 1e2k HIS58 GLU83 TRP88 ILE97 ILE ChiRotor calculation with ligand present RMSD to Crystal Structure side chain coordinates TYR101 GLN125 MET128 TYR132 ARG163 His164 TYR Side chains within 4 angstroms from any ligand atom selected for refinement Blank means side chain was not selected for refinement ARG ARG < 2.0 GLU >

11 ChiRotor: Placement of Side Chains in Thymidine Kinase Structures Complex 1e2k RMSD (Å) 1.09 Time (secs) 176 ChiRotor calculation with ligand present 1e2m 1e2p 1ki2 1ki3 1ki RMSD to Crystal Structure side chain coordinates Side chains within 4 angstroms from any ligand atom selected for placement 1ki6 1ki7 1kim Computing time for Dell M 70 Pentium 2 GHz single processor 1qhi ki ki

12 Side Chain Conformations in HUMAN CDK 2 CMPLEXED WITH THE INHIBITR STAURSPRINE -ChiRotor without the ligand 1aq1 ILE VAL LYS VAL PHE GLU PHE LEU HIS GLN ASP GLN ASN LEU ASP Protein shown with Ligand

13 ChiFlex FlowChart Protein 3D Structure Select a Set of n Residues For Flexibility testing Remove all side chain atoms of selected residues Start loop for i from 1 to n Choose Residue i Sample side chain conformations of residue i and energy minimize using CHARmm Accept conformation if pass energy threshold and degeneracy checking End loop for i Construct complete structure by making all possible combination of the partial structures and energy minimize using CHARMm utput: multiple conformations with varing side chain conformations of the selected residues utput: a set of partial structures 13

14 Identification of Flexible Residues Thymidine Kinase (1kim) His58 Glu83 Ile97 Ile100 Gln125 Arg163 Tyr172 Arg176 Arg222 Glu225 Lowest RMSD to Crystal Side Chains ~1.0 angstroms CDK2 (1aq1) Ile10 Lys33 Phe80 Glu81 Leu83 His84 Gln 85 Asp86 Gln131 Asn132 Leu134 Asp145 Lowest RMSD to Crystal Side Chains ~1.8 angstroms Residues identified by ChiFlex as highly flexible (>2 Å) 14

15 Energy of Diverse Protein Side chain conformations Relative Energy nsc 1a4q 1aq1 1dm2 1err 3ert 1rth 1c1c 1kjo No of Conformations 15

16 Flexible Docking Protocol Generate Protein Side Chain conformations using ChiFlex Generate Ligand Conformation using CatConf /CAESAR Compute Hotspots Dock to Hotspots Retain n poses LibDock Modify Side Chain conformations using ChiRotor Anneal/ Energy minimize Ligand pose CDCKER 16

17 New Flexible Docking Protocol CDCKER 17

18 Step 1: Generate Receptor Conformations Receptor Generate n Receptor Structures Made up of different Side Chain Conformations (ChiFlex) Generate reasonable low energy side chain conformations Side-chain/backbone interactions explicitly taken into account Flexible residues specified by user and protocol Need to run only once per receptor binding site 18

19 Step 2: Compute Protein Hotspots Receptor Generate n Receptor Structures Made up of different Side Chain Conformations (ChiFlex) Compute Hotspots on Conformation n Simple Simple two-feature model model has has been been shown shown to to be be accurate and and robust robust for for guiding docking 1 1 Red = polar hotspots Blue = apolar hotspots Diller and Merz,. PRTEINS: Structure, Function and Genetics 43: (2001)

20 Step 3: Generate Small Molecule Conformations Receptor Generate n Receptor Structures Made up of different Side Chain Conformations (ChiFlex) Compute Hotspots on Conformation n Small Molecules Generate Small Molecule Conformation (CatConf ) Several Several Methods Methods Available: Available: ptimize ptimize either either speed speed or or accuracy accuracy Generates diverse low energy conformers Fast (~ 50 compounds per minute) N H NH 2 NH 2 NH 2 NH 2 20

21 Step 4: Fast and Efficient Docking to Hotspots Receptor Generate n Receptor Structures Made up of different Side Chain Conformations (ChiFlex) Small Molecules Generate Small Molecule Conformation (CatConf ) Compute Hotspots on Conformation n N H NH 2 NH 2 NH 2 NH 2 Dock to Hotspots (LibDock) Retain m Poses Triplets of hotspots matched to triplets of small molecule atoms Matches clustered and optimized Knowledge-based docking: only only sample relevant regions of of binding site site 21

22 Step 5: Side-Chain ptimization Around Docked Pose Receptor Generate n Receptor Structures Made up of different Side Chain Conformations (ChiFlex) Small Molecules Generate Small Molecule Conformation (CatConf ) Compute Hotspots on Conformation n Dock to Hotspots (LibDock) Retain m Poses N H NH 2 Refine Side Chain Conformations (ChiRotor) CHARMm-based refinement of side chain positions in the presence of docked molecule pose 1-3 minutes per pose 22

23 LibDock: Fast and Accurate Docking Solution Docking results on the newly available AstexDiverse 1 dataset 85 receptor-ligand PDB entries comprising diverse receptor families Docking under a minute per small molecule RMSD bin (Å) LibDock % Docked Successfully 1 CatConf BEST CAESAR 2 <1 61% 67% <2 91%. 86% Success Rate 91% 86% % of AstexDiverse dataset docked successfully with LibDock using two conformation generation methods Best of 10 highest-ranked poses 2. Li et al, submitted to J. Chem. Inf. Model

24 Ligand Pose, i CDCKER Flowchart Receptor Calculate Grid Generate Conformations (high temp MD), n i*n Generate rientation (m) (in the presence of Rec) And check vdw E Threshold i*n*m Simulated Annealing i*n*m Minimization i*n*m Rank Best Poses, j i*j 24

25 CDCKER Results: RMSD to 41 Crystal Structures RMSD apu (PENICILLPEPSIN ) RMSD uvs (Thrombin) RMSD 1htf (hiv protease) 1pgp (Phosphogluconate Dehydrogenase ) Proteins 25 Same Dataset as Erickson et al. J Med Chem (2004) 47:45-55

26 Docking Solutions ptimized for Speed or Accuracy Comparative performance of LibDock and CDCKER on AstexDiverse 1 dataset Method % Docked Accurately (Best) 2 % Docked Accurately (Top) 3 RMSD Average (Best) RMSD Average (Top) Time (s) 4 LibDock CDCKER LibDock LibDock is is optimized optimized for for speed: speed: : : get get accurate accurate docked docked poses poses in in seconds seconds CDCKER CDCKER is is optimized optimized for for accuracy: accuracy: : : get get significant significant improvement improvement in in rankordering rankordering of of correct correct pose pose and and RMSD RMSD to to X- X- ray ray structure structure Hartshorn, et al. J. Med. Chem., 50 (4), (2007) 2. Best of 10 highest-ranked poses 3. Top-ranked pose 4. 3 GHz Intel Xeon, Linux

27 Flexible Docking of 1kim ligand into 1kim receptor Xray Ligand is shown in green. Ligand RMSD to xray = 0.8 angstrom 27

28 Flexible Docking of 1ki4 ligand into 1kim receptor Xray Ligand is shown in green. Ligand RMSD to xray = 1.0 angstrom 28

29 Validation of Rational Flexible Docking Protein Thymidine Kinase Estrogen Ligand 1kim 1ki4 1err Receptor 1ki4 1kim 3ert Flexible Receptor Lowest RMSD Rigid Receptor RMSD ert 1err CDK2 1aq1 1dm dm2 1aq vyw 1dm CX2 1cx2 3pgh pgh 1cx Neurominidase 1nsc 1a4q Cross-docking: Dock ligands into an alternate conformation of the same receptor HIV-RT 1a4q 1rev 1s1x 1fk9 1s1x 1nsc 1rth 1rth 1rth 1rev rth 1rev 1.54 RMSD Values compared to X-ray conformation for cross-docking experiments 1fk9 1rth 1rth 1s1x 1s1x 1s1x 1fk9 1fk fk9 1rev 0.35 Thermolysin 1kr6 1kjo kjo 1kr

30 Flexible Docking: Unprecedented Customizability Loop Sampling (Looper 1 ) Generate Protein Side Chain Conformations (n) (ChiFlex) Generate Small Molecule Conformations (CatConf ) Use Pre-generated Conformations (e.g.: MD Trajectory Frames) Compute Hotspots on Conformation n Easily change workflow to incorporate different methods Add CHARMm-based loop sampling 1 in the beginning Change the docking engine Dock to Hotspots (LibDock) Retain m Poses Refine Side Chain Conformations (ChiRotor) Energy minimize m Poses (CHARMm) Replace with LigandFit CDCKER GLD Refine docked Poses (CDCKER) V. Z. Spassov, L. Yan, P. K. Flook, to appear in Protein Science

31 Improvements in DS 2.0 Flexible Docking protocol based on LibDock technology Ligand Conformational sampling by CatConf or CAESAR Improved CDCKER Ability to handle structures with cofactor Parallel Deployment of the protocol Improved user interface (input parameters/utput data) 31

32 Summary of Accelrys Flexible Docking Method A rational approach Consistent force field (CHARMm) used throughout Minimal user intervention required Docking into a realistic environment Docking of ligand influenced by existing side chains in active site Accurate initial protein conformations Realistic and based on side-chain-backbone interactions as well as side-chain-side chain interactions Pre-generated side chain conformations can be used as input Library of receptor conformations can be saved and used for all small molecules being docked Customizable Parellel Execution in multi-processors and clusters 32

33 Strengths of the Approach ChiRotor ability to reliably predict side chain conformations Very flexible interface Power of modifying the workings by modifying workflow Extendable interface adding LPER Docking engine may be replaced Consistent, validated force field CHARMm A rational approach to Ligand Protein interaction problem 33

34 Thanks to Accelrys R&D Al Maynard (Structure Based Design) Eric Yan (CHARMm) Jürgen Koska (Flexible Docking Workflows) Lisa Yan (Proteins) Nan-Jie Deng (Simulation, CHARMm) Velin Spassov (ChiRotor, ChiFlex) Paul Flook (Direction) Nic Austin (Direction) Frank Brown (Discussions) Marketing Sylvia Tara, Dipesh Risal Application Scientists ver 25 PhD scientists External Collaboration Charles E. Brooks (CHARMm, CDCKER) Dave Diller (LibDock) 34

35 Future DS 2.0 Webinars ctober 9, 2007 ctober 11, 2007 ctober 16, 2007 November 1, 2007 November 6, 2007 November 8, 2007 November 20, 2007 verview of Discovery Studio 2.0 >>> New Simulation Methods for Drug Discovery: Fast Accurate pk Prediction and More >>> A Rational Approach to Receptor-Flexible Docking: Method and Validation Results >>> Going where no Pharmacophore has Gone Before: Fragment-based design and activity profiling >>> Practical QSAR and Library Design: Advanced Tools for Research Teams >>> Antibody Modeling: New Tools for Improved Success >>> Elegant Protein-Protein Docking in Discovery Studio >>> 35

36 Accelrys Science Forums Come see Discovery Studio in action at the US and European Science Forums: Paris, France 18 ctober Boston, MA 23 ctober Princeton, NJ 25 ctober San Diego, CA 30 ctober Heidelberg, Ger 31 ctober Cambridge, UK 13 November San Francisco, CA 15 November Agenda, Registration details, etc can be found at: 36