Dipesh Risal, Ph. D. Life Sciences Product Manager June 26 ...media.accelrys.com/webinars/DS-21-Series/6-26-DR-QMMM.pdf · presentation and related documents, whether oral or written,

Towards Increased Accuracy in

Computational Drug Discovery with QM/MM

Towards Increased Accuracy in

Computational Drug Discovery with QM/MM

This presentation and/or any related documents contains statements regarding our plans or expectations for future features, enhancements or functionalities of current or future products (collectively "Enhancements"). Our plans or expectations are subject to change at any time at our discretion. Accordingly, Accelrys is making no representation, undertaking no commitment or legal obligation to create, develop or license any product or Enhancements. The presentation, documents or any related statements are not intended to, nor shall, create any legal obligation upon Accelrys, and shall not be relied upon in purchasing any product. Any such obligation shall only result from a written agreement executed by both parties. In addition, information disclosed in this presentation and related documents, whether oral or written, is confidential or proprietary information of Accelrys. It shall be used only for the purpose of furthering our business relationship, and shall not be disclosed to third parties.

Dipesh Risal, Ph. D.Life Sciences Product ManagerJune 26, 2008

Dipesh Risal, Ph. D.Life Sciences Product ManagerJune 26, 2008

2© 2007 Accelrys, Inc.

OverviewOverview

• QM/MM background• QM/MM implementation• Validation: Scientific applications of QM/MM


QM/MM overviewQM/MM overview

• Combine QM and MM methods to achieve good accuracy at low cost– Treat ‘chemically

significant’ region with QM– Treat the bulk with MM– Combine the results

QM region

MM region

Ligand


QM/MM in Discovery StudioQM/MM in Discovery Studio

• A QM/MM method has been implemented that incorporates– DMol3 (DFT) for the QM

region– CHARMm for the MM region– QUANTUMm, a

communication program between the two regions

• How do each of these programs work?

QM region

MM region

Ligand


QM/MM OverviewQM/MM Overview

Anatomy of QM/MM machinery

• QM and MM servers only provide energies and gradients• Propagation algorithms (geometry optimization, MD, TS search ...) operate on data received from compute engines

QM/MM Driver

QM Engine

xyzE∇Eq

QM/MM Driver

MM Engine

xyzq

E∇E

QM/MM Driver

Call QM Call MMGeometry

optimization or MD step

Repeat


DMol3 OverviewDMol3 Overview

• DMol3 uses DFT to predict structures, energies, electronic properties

• Works for molecular and periodic systems• Extremely fast

– Numerical basis sets provide a rapid means of evaluating Coulomb and exchange-correlation potentials

– Provides options to trade off between computational cost and accuracy

• Delley, J. Chem. Phys.113, 7756 (2000)– Energy calculations on drug-size molecules require few

minutes on typical laptop


DMol3 FunctionalsDMol3 Functionals

H-bonded complexes Dispersion-dominatedcomplexes

Mixedcomplexes

Type of complex DMol3/PBE DMol3/HCTH Z-T B3LYP1 Z-T PBE1

H-bond 1.0 3.4 2.0 1.3

Dispersion 3.2 2.8 6.5 4.9

Mixed 1.2 0.9 2.9 1.9

Avg. Error (kcal/mol) of binding energies in loose complexes

1. J. Chem. Theory Comput. (2007), 3, 289-300.


MM engine: CHARMmMM engine: CHARMm

• CHARMm: the industry standard for simulation of macromolecules and protein-ligand systems• Constant Forcefield development

– Alex MacKerell, Charlie Brooks, Bernard Brooks, Accelrys, Others• Most comprehensive simulation package available

– MM, MD– CDOCKER– Free Energy Perturbation (FEP)– MM-PB/GB SA Scoring– Normal Mode analysis– RDOCK (refinement of Protein-Protein docking)– ChiRotor, Looper– Replica Exchange Molecular Dynamics (REMD)– Three implicit membrane models

• GBSW, GBSA-IM, IMM1

– Umbrella sampling– Monte Carol simulations– Physics-based pK Prediction and Protein Ionization– Constant-pH MD– And many more!

• Strong UI support in Discovery Studio– Antibody Modeling– Implicit Membrane Modeling – Receptor-Flexible Docking


QUANTUMm OverviewQUANTUMm Overview

• Scientific requirements for biological QM/MM:– QM region “feels” MM atom

environment– Minimal number of FF parameters

for QM region• Issues to address

– Embedding: how does the QM region interact with the MM region?

– Boundary region: what happens to bonds that cross between regions?

),( QMMMtotal RREE =

• Total energy ...

)()()(),( / IOEIEOEIOE MMQMQMMMtotal ↔++=


QM ↔ MM Coupling: ElectrostaticsQM ↔ MM Coupling: Electrostatics

MM → QM

QM → MM

QM density is polarized by MM point charges:electronic embedding

O charges polarize I density

I gradients induce forces on O

Part of QM energy expressionelecMMQME /


QUANTUMm Issue: Broken QM ↔ MM bondsQUANTUMm Issue: Broken QM ↔ MM bonds

→

add link atom (L) to QM calculation

Problem: QM calculation on I region yields unrealistic species

• Link atom is absent in MM calculation• Position restrained onto CA -CB vector• QM/MM server program handles link atoms transparently

[Field, JCC 1990]


Complications: ElectrostaticsComplications: Electrostatics

Electronic embedding and link atoms

Problem: QM overpolarization near link atoms: MM host too close

Solution: the QM fragment should see no charge from MM host atom

[Bakowies, JPC 1996; Sinclair, J Chem Soc Faraday 1998]


QUANTUMm User InterfaceQUANTUMm User Interface

• DS QUANTUMm allows easy job setup • Select QM region• Set DFT options

– Charge on QM region– Spin multiplicity of QM region– Relative precision (basis set, integration grid, SCF

convergence)

• Number of processors for parallel calculation• Provides the tools you need to balance the size of the

problem, relative accuracy, and computational cost


QM/MM Implementation IQM/MM Implementation I

• Methods available for first release– QM/MM Minimization – QM/MM Energy Calculation (Single Point Energy)

• If only the ligand is in the QM region, both protocols output ESP, Hirshfeld and Mulliken charges for ligand

– Recharge Ligand Pipeline Pilot Component• Point charges from a protein model used in the electronic

structure calculation, causing polarization of the ligand


QM/MM Calculation: Setup in Discovery StudioQM/MM Calculation: Setup in Discovery Studio


QM/MM Applications for Life Science QM/MM Applications for Life Science

• QM/MM-derived ligand partial charges for:– improved docking accuracy (and simulation)

• Optimization of hydrogen bonds – Post-processing of docked poses

• Modeling of special electrostatic interactions not fully captured by force fields

– Cation-Pi interactions– Pi-Pi interactions– Charge transfer – Metal-ligand-protein interactions

• Improved estimate of interaction energy between protein and ligand

• QM/MM in conjunction with MD1, QM-PBSA2

• Preparing ligands and cofactors for MM calculations3

– Heme, others

• Studying reaction mechanisms3

– individual steps (hydrolysis, etc.)– transition state– entire catalytic cycle

• Activation free energy3

• QM/MD (dynamics)• Semi-empirical method in QM/MM

1. J Comput Aided Mol Des. 2007 Jan-Mar;21(1-3):131-72. J Phys Chem B, 109 (2):10474-83. 3. Chem Rev. 2006 Sep;106(9):3497-5194. Drug Discov Today: Technologies, 2004 Dec; 1(3), 253-2605. Drug Discov Today. 2007 Sep;12(17-18):725-31 .

Possible Today. Some validationAlready available

Possible Today. Validation ongoing

Possible in future releases based on customer prioritization


QM/MM Background: Partial Charge AnalysisQM/MM Background: Partial Charge Analysis

MM Min QM/MM Min

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

C11 O11 O12 C10 C9 C8 C7 C2 S1 C6 C5 N1 C3 O3 N2 C4 H1 H2 H3 H4 H5 H6 H7 H8 H9H10 H11 H12 H13 H14 H15 H16

CHARMm M-Rone

CFF

QM/MM Min

Generally, H-bond donors become increase in δ+H-bond acceptors increase in δ-

Partial charges on ligand after MM and QM/MM Minimization, 1STP-1: Streptavidin/Biotin


QM/MM Applications: BackgroundQM/MM Applications: Background

• CDOCKER is a CHARMm-based small molecule docking refinement algorithm1

– Uses soft-core potentials– Grid-based (optional)

1. Wu et al. J Comput Chem (2003) 24:1549-62

Generate ligand conformationsthrough high temperature MD

(grid-based) simulated annealing

Full minimization

Output # of refined ligand posessorted by energy

(vdW+ e/s + ligand strain)

Random (rigid-body) rotation


QM/MM Applications: Improving Docking Accuracy*QM/MM Applications: Improving Docking Accuracy*

• CDOCKER on AstexDiverse Dataset** (85 diverse, high-resolution protein-ligand complexes)

Top Pose

Dock with CDOCKER

CDOCKER

Calculate RMSD to X-ray

Prepared X-rayProtein- Ligand Complex

CDOCKER-QM-CDOCKER

Calculate QM charges for ligand

CDOCKER

Top Pose

Dock with CDOCKER

Calculate RMSD to X-ray

Top Pose

Randomize Ligand Conformation

1.02891.6331Avg.RMSD

88.10%72.62%Success

Best Pose

First Pose

X-Ray (A)

1.01.6 Avg.RMSD

88%73%Success

Best Pose

First Pose

X-Ray (A)

1.02891.6331Avg.RMSD

88.10%72.62%Success

Best Pose

First Pose

X-Ray (A)

1.01.6 Avg.RMSD

88%73%Success

Best Pose

First Pose

X-Ray (A)

0.91141.3856Avg.RMSD

90.48%77.38%Success

Best Pose

First Pose

X-Ray (B)

1.31.4Avg.RMSD

88%79%Success

Best Pose

First Pose

X-Ray (B)

* Cho et al, J Comput Chem 26: 915–931, 2005 ** Hartshorn et al. JMC 2007

Poses are scored with CDOCKER Energy: sum of electrostatics+vdW interaction energy + ligand strain E


QM/MM Applications: Improving Docking AccuracyQM/MM Applications: Improving Docking Accuracy

•

Available options: ESP, Hirshfeld, Mulliken

Available Options: Coarse/Medium/Fine.Affects the basis set, k-point, and SCF convergence criteria

Available Options: local (LDA) potentials (PWC, VWN) and gradient- corrected (GGA) potentials (PW91, BP, PBE, BLYP, BOP, VWN-BP, RPBE, HCTH).

• Pipeline Pilot workflow for CDOCKER-QM-CDOCKER


QM/MM Applications: Improving Docking AccuracyQM/MM Applications: Improving Docking Accuracy

• CDOCKER-QM-CDOCKER on AstexDiverse dataset: success by RMSD bin

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

< 0.5 Å < 1.0 Å < 1.5 Å < 2.0 Å

CDOCKER First PoseQM-CDOCKER First Pose

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

< 0.5 Å < 1.0 Å < 1.5 Å < 2.0 Å

CDOCKER Best PoseQM-CDOCKER Best

Only single, top-rankedpose for each PDB complex considered

Best-RMSD pose (to X-ray)out of 10 docked poses considered


QM/MM Application: Cation-Pi Interaction ModelingQM/MM Application: Cation-Pi Interaction Modeling

• Investigate the binding of Compound 1(active against Histamine H3 receptor) on AChE1

• Goal was to design dual-acting compound• Hypothesis of two cation-pi interactions

between ligand and receptor

• Docking (CDOCKER) of Compound 1 into AChE receptor (PDB ID 1EVE) yielded a pose in position to make cation-Pi interactions

• Cation-Pi not modeled well by force fields

• QM/MM geometry optimization suggests cation-Pi interactions

– 51 hrs on 2 processors– Multiplicity: Smart– Quality: Ultra Coarse– Infinite nonbonded cutoffs– No constraints/restraints

1. Bembenek, S. D.. et al., Bioorg. Med Chem. (2008), doi:10.1016/j.bmc.2007.12.048

Einitial = -15676.278608 kcal/molEfinal = -16904.454283 kcal/mol


QM/MM Application: Optimizing Heme systemsQM/MM Application: Optimizing Heme systems

• Starting Structure: PDB ID 1P2Y1, Cytochrome P450CAM in complex with (S)-(-)-Nicotine

1. Biochemistry 42: 11943-11950

• improved coordination between Arg and heme carboxyl groups• coordination of Fe (II)

Arg 112

Cys 357

Arg 299

His 355

nicotine

Arg 112

Cys 357

Arg 299

His 355

nicotine

QM/MM min.

2 proc. 2.6 GHz Xeon ca. 4 days


QM/MM Application: Optimizing Heme systemsQM/MM Application: Optimizing Heme systems

• Starting Structure: PDB ID 1P2Y1, Cytochrome P450CAM in complex with (S)-(-)-Nicotine

1. Biochemistry 42: 11943-11950

Initial QUANTUMm Energy = -29288.570302 kcal/mol

Initial QM Energy = -13252.109685 kcal/mol

Initial MM Energy = -16036.460617 kcal/mol

QUANTUMm Energy = -33719.154699 kcal/mol

QM Energy = -14279.192571 kcal/mol

MM Energy = -19439.962129 kcal/mol

6-coordination of Zinc in QM/MM optimized structureNo constraints/restraints were used in the QM/MM optimization experiment


QM/MM Application: Modeling Metals in ProteinsQM/MM Application: Modeling Metals in Proteins

From Jain et al , PROTEINS 2007

ligand bound to zinc ion

development of zinc force field difficult

Zinc Metalloproteins: important drug targets

coordination patternelectrostatics vdW interactions

QM/MM description challenging

large QM zoneQM/MM bonds


QM/MM Application: Optimization of MMP Binding SitesQM/MM Application: Optimization of MMP Binding Sites

• Previous study1 describes design and docking studies with matrix metalloproteinase-9 (MMP-9) and a set of hydroxamate inhibitors

• The study was replicated in Discovery Studio 2.1– Sketch “Compound 1”– Dock to MMP9 receptor (PDB ID 1GKC) with CDOCKER

1. J. Med. Chem, 2005, 48, 5437-54472. J. Med. Chem. 2002, 45, 919-929

Known hydroxamate interactions in MMP-9 binding site2

Top docked pose using CDOCKER




• The study was replicated in Discovery Studio 2.1– Sketch “Compound 1”– Dock to MMP9 receptor (PDB ID 1GKC)– Take First Pose, optimize protein-ligand complex with QM/MM




• Some details of QM/MM experiment– 3 His, Zinc, Glu402, Ligand included in QM region– Cut between QM and MM region made at the CA-CB bond of residue

side chains– 5Å shell around QM region subjected to MM (CHARMm)– Rest of the system kept frozen (as per published study1)– 2000 steps of minimization, PBE Functional, Ultra-Coarse setting

(Basis Set: minimal, Integration Grid: xcoarse, DMol3 Cutoff 3.0 Å, SCF Density Convergence 5.0e-4)

– 7 hours on 4 CPU 1.8GHz Opteron machine

1. J. Med. Chem, 2005, 48, 5437-5447






Before QM/MM (Docked Pose) After QM/MM

Glu402Glu402

Zn

His411

His405

His401 Zn

His411

His405

His401






QM/MM optimized distances and partial charges in a hydroxamate-MMP9 complex from a published study1

Glu402

Zn

His411

His405

His401



• Comparison of post-optimized interactions of two MMP-9 actives– Ongoing work: QM/MM-based scoring and rank-ordering of actives in the series1,2


Ki = 5.05 nM Ki = 0.08 nM


Glu402

Zn

His411

His405

His401

Glu402

Zn

His411

His405

His401


Glu402

Zn

His411

His405

His401

Glu402

Zn

His411

His405

His401

QM/MM Inter E =-246.95 Kcal/mol

QM/MM Inter E =-266.92 Kcal/mol

“Compound 1” “Compound 20”


Summary/ Future PlansSummary/ Future Plans

• QM/MM methods have been shown to provide improvement over pure force field (CHARMm) calculations

• QM/MM methods have been applied in real-life computational tasks– Improved partial charges for docking– Modeling of special interactions such as cation-pi– Refinement of heme-containing systems– Optimization of metalloprotein active sites

• accurate interaction energies• Ongoing validation

– QM/MM-based scoring function– Comparison of DMol3 ESP charges with AM1-BCC, others– Torsion profiles (ΔE vs. torsional angle) for select small molecules

• Future developments on QM/MM will be exclusively based on customer feedback– QM/MM based scoring function– Semi-empirical methods for QM– Modeling reaction mechanisms


Thank you!!!Thank you!!!

• Thank You for attending today’s webinar. If you have any further questions please e-mail me at: [email protected]

• You can also contact us using the form on our website: http://accelrys.com/company/contact/

• We will be exhibiting at the following upcoming events:– CHI Protein Kinase Targets (June 23 – 25, Boston, Booth #4)– CHI Structure Based Design (June 25 – 27, Boston, Booth #7)– Drug Discovery Technology and Development (August 4 – 7, Boston, Booth #512)– ACS Fall 2008 (August 17 – 21, Philadelphia, Booth #211)

• Reminder: the next webinar in this series will be:

– Pharmacophore Guided Fragment-Based Drug Design Dr. Tien Luu – July 10, 2008 at 7am PST and 10am PST

http://accelrys.com/company/contact/


QM ↔ MM coupling: Short-rangeQM ↔ MM coupling: Short-range

•QM ↔ MM van-der-Waals interactions handled classically (i.e., by MM server)

•Requires Lennard-Jones parameters for QM region


HOMO/LUMO Visualization (Work in Progress…)HOMO/LUMO Visualization (Work in Progress…)

Diels-Alder Reaction

The electron-rich HOMO of the diene and the Electron-vacant LUMO of the dienophile mustbe in a stacked orientation (top-bottom) formaximal overlap of the orbitals for the reactionto proceed

Diene Dienophile

HOMO LUMO

Documents

Dipesh Risal, Ph. D. Life Sciences Product Manager June 26 ...media.accelrys.com/webinars/DS-21-Series/6-26-DR-QMMM.pdf · presentation and related documents, whether oral or written,