per-ola norrby december 2014 molecular modeling in pharmaceutical development

Per-Ola NorrbyDecember 2014

Molecular Modeling in Pharmaceutical Development


Pharma Modeling

Discovery phase

Screening candidate molecules• thousands-millions• pharmacokinetic properties• binding to target

QSARDocking…

Product Development phase

Single moleculesProcessability• Crystal propertiesDegradation• Chemical reactivitySynthesis• Reagent selection• Reaction conditions


Chemical synthesis

PotencySolubility

Permeability

MetabolismPatentability

Stability

Distribution

Toxicology

Processability


Discovery Modeling

Fast methods, many molecules

QSAR, > 106

Docking, molecular mechanics, > 103

MD, 1-100

Fast

Medium

Slow

Candidates

Enriched set


Product Development Modeling

Accurate methods, single molecules, devices

Synthesis route, degradation possibilities Databases of reactions, ICSynth, Zeneth

Synthesis or degradation predictions QM, DFT, MM

Material properties Crystal packing, periodic boundaries, MM, DFT

Distribution, shelf life Kinetics networks, simulation

Flow dynamics Finite elements


Quantum Chemical Reactivity

Calculate all possible species on the path

Calculate the transition states connecting them

BrHN

O

Pd(PtBu3)2

tBuOKN O+

Example: Buchwald-Hartwig reaction



Pd BrtBu3P

Pd NtBu3P O

Pd BrtBu3PN

O

H

PdtBu3PN

O

HOtBu

PdtBu3PN

O

H

Pd OtButBu3PN

O

H

Pd OtButBu3P

+200 kJ/mol

G

NPdtBu3P

‡

O

Sunesson, Limé, Nilsson Lill, Meadows, Norrby, JOC 2014, ASAP



Simplified!

Pd OtButBu3P

N

O

H

NPdtBu3P

‡

O

�G‡ =

79

kJ/m

ol

Clot, Norrby, inInnovative Catalysis in Organic Synthesis: Oxidation, Hydrogenation, and C-X Bond Forming Reactions, Wiley, 2012


DBU as base in nonpolar solvent

Formation of DBU-H+ Br-

Very costly in non-polar solvent


Bases in polar solvent (DMF)

t-BuO– DBU


Modeling reaction selectivity

G‡

Only TS needed


The High-Energy Intermediate

Selectivity

Approximate selectivity

OAcR

NuR

PdLL

R

Oslob, Åkermark, Helquist, Norrby, Organometallics 1997, 3015



Approximate selectivity

Liljenberg, Brinck, Herschend, Rein, Rockwell, Svensson, J. Org. Chem. 2010, 4696

R

R

Br

R

Br

H



Autoxidation

R H

R

R OO


The Autoxidation tool

R H

RR OO

R OOH

O2

Autoxidation propagation

Bond dissociation energy < 370 kJ/mol R H R H+H

2D-structure 3D-structure

Remove H B3LYP opt

UB3LYP opt

BDE calc.

Andersson, Broo, Evertsson,J. Pharm. Sci. 2014, 103, 1949


Modeling reaction selectivity

G‡

Boltzmann summation1-10 paths

>10 000 possible!


Experimental Optimization

Objectives

ReactivityStabilitySelectivity

Factors

TemperatureH2 PressureSolventBase?Additives?Ligand(s)

Design Points

2-32-3

3-100-50-5

8-500

96 >1 000 000

Optimization space discontinuous – Per Ryberg, SYNFLOW meeting 2011– Dense testing needed


Modeling Methods

Needed: transition states (TS) for all pathways

Quantum Mechanics, DFT + Accurate − Time consuming, hours-days for each TS − TS search not automated

Molecular Mechanics + Fast, seconds or less per structure + Adjustable parameters + Automated conformer search ? Accurate ? (target: 2 kJ/mol) − Not available for metals, TS


Reaction specific force fields

Quantum Mechanics (QM) Molecular Mechanics (MM)

Q2MM

Norrby, Brandt, Rein, J. Org. Chem. 1999, 5845


Asymmetric Dihydroxylation

N

MeO

N

OR

OH

OHcat. OsO4

NMO or K3Fe(CN)6

MUE: 2.5 kJ/mol

Norrby, Rasmussen, Haller, Strassner, HoukJ. Am. Chem. Soc., 1999, 10186.

Fristrup, Jensen, Andersen, Tanner, NorrbyJ. Organomet. Chem. 2006, 2182.

1 10 100 1000 100000.3

3.0

30.0

300.0

3000.0

30000.0

Exp. e.r.

Ca

lc. e

.r.


0000

0000

0000

00000

0000

8

X

0000

0000

0000

00000

0000

8

1

Parameterization from QM data, structures, charges, relative energies.Vibrational data (Hessians) is modified before being used as reference data:

Diagonalization:

Eigenvalue replacement:

Forming the new Hessian:

Q2MM Norrby, J. Mol. Struct. (Theochem) 2000, 506, 9


The Q2MM ProcedureQuantum mechanical

transition states

Structural data(e.g., bond lengts)

Vibrational data(force constants)

Modification of reaction coordinate (minimum)

Set ofQM data

Select force field(MM3*, Amber)

Define new atom& bond types

Guess initialparameters

Generate trial force fields

Sets ofMM data

Comparesets

Bestforce field

Lowest error

Finalforce field

Optimum?

Test againstexperiments


Q2MM

QM TSca 20 atoms

MM TSFF

>100 atoms

Conf. Search∆∆E‡Boltzmann

Eyring, e–∆∆G‡/RT

Exp. ratio


Q2MM for AD: Development

B3LYP for small models (ca 50, different R).Haller, Strassner, Houk, J. Am. Chem. Soc. 1997, 119, 8031.

Define Q2MM force field for the region around the reaction center.Fit structures, energies, vibrations, charges to the B3LYP results for all small models.

Use Q2MM model for real system(ca 1/day/cpu)


Asymmetric Dihydroxylation

N

MeO

N

OR

OH

OHcat. OsO4

NMO or K3Fe(CN)6

MUE: 2.5 kJ/mol

Norrby, Rasmussen, Haller, Strassner, HoukJ. Am. Chem. Soc., 1999, 10186.

Fristrup, Jensen, Andersen, Tanner, NorrbyJ. Organomet. Chem. 2006, 2182.

1 10 100 1000 100000.3

3.0

30.0

300.0

3000.0

30000.0

Exp. e.r.

Ca

lc. e

.r.


Asymmetric Hydrogenation

Donoghue, Helquist, Norrby, Wiest, J. Am. Chem. Soc. 2009, 410

0.001 0.01 0.1 1 10 100 1000 100000.01

0.1

1

10

100

Exp. e.r.

Ca

lc. e

.r.

MUE: 3.2 kJ/mol


Ru-catalyzed ketone reduction

MUE: 2.7 kJ/mol

0.001 0.01 0.1 1 10 1000.001

0.01

0.1

1

10

100

1000

Exp. e.r.

Ca

lc. e

.r.

Limé, Lundholm, Forbes, Wiest, Helquist, Norrby, J. Chem. Theory Comput. 2014, 2427


Natural Q2MM

New:Eigenvector-projected Hessian

DifferentiateEigenvalue fit from Eigenvector fit

Fit w/o lowest mode

Limé, Norrby,

J. Comput. Chem. 2014, in press


Q2MM at AZ

What will the user do?

Existing reaction: New reaction: – Draw substrate – Convince us to develop it – Select reaction – Wait (months) – Push GO – Go to “Existing reaction”

Where are we now?• Manual screening of ligands, few reactions

• Experimental testing under way (Macclesfield)• Tools for implementing new reactions

Under development (with Notre Dame University)• Improved, faster tools for new reactions• Automated Virtual Screening (Target: March 2015)

Elaine Limé


AcknowledgmentDenmarkTorben RasmussenPeter FristrupMårten AhlquistSigne T. Henriksen

Notre DamePaul HelquistOlaf WiestPatrick DonoghueElsa KiekenAaron ForbesEric Hansen

$$$The University of Gothenburg, The Swedish Research Council,AstraZeneca, FP7/SYNFLOW, COST, C3SE/Gothenburg

AstraZenecaTobias ReinPer RybergHans-Jürgen FederselRobert WoodwardRebecca MeadowsSten Nilsson LillSimone TomasiElaine LiméAnders BrooRachel MundayDavid Buttar

U. GothenburgAnna HedströmPer-Fredrik LarssonJonatan KleimarkCharlotte JohanssonPetra RönnholmCarina BäcktorpYlva Sunesson

LundUlf RydePatrik Rydberg

per-ola norrby december 2014 molecular modeling in pharmaceutical development

Documents

asapperola norrbydecember

ts ts search

pathways quantum mechanics

transition states ts

automated molecular

modeling methodsneeded

pharmaceutical development

molecules qsar