Design of Experiments (DoE): A “New” Approach to Reaction ... · Design of Experiments (DoE): A “New” Approach to Reaction Optimization Steven A. Weissman NEACS Symposium:

Confidential

Design of Experiments (DoE): A “New” Approach to Reaction

OptimizationSteven A. Weissman

NEACS Symposium: Oct 22,2010

OutlineBasic Principles of Design of Experiments (DoE)Case StudiesTake Home message/Questions

Confidential

Current Approach to OptimizationChange One Factor at a time (OFAT)

o Rarely uncovers the optimal conditions• Local vs global maxima

o Different conclusions depending on starting pointo Requires many experiments/little informationo Cannot separate “noise” from true variability

• Is a 2% yield gain real or just run-to-run variation

Confidential

Current Approach to OptimizationChange One Factor at a time (OFAT)

o Rarely leads to optimal conditionso Leads to different conclusions depending on starting pointo Requires many expts/little informationo Cannot separate “noise” from true variability

o Ignores interactions of variables

Confidential

OFAT Example: 2007

Confidential

OFAT: 21 Reactions

Confidential

DOE vs OFAT

OFAT: 3 factors needed 21 reactionso No information on interactions of effectso No information on robustness; near ‘edge of failure’

DOE: 3 factors: 11 reactionso Better quality informationo Learn about interactions of effectso 10 Fewer reactions

Confidential

Notable Quote

“If you test one factor at a time (OFAT), there’s a low probability that you are going to hit the right one before everybody gets sick of it and quits”

Forbes magazine article on DOE (1996)

Confidential

What is DOE ?• Statistically-based set of expts in which all chosen factors are

varied simultaneously• ‘Continuous’ factors are ideal (time, temp, equiv)

• the ‘How much/many’

• Analysis reveals which factors influence the outcome and identifies optimal conditions

• Systematic, organized approach to problem solving• Generates a mathematical model of the design space• Integral component in the QbD movement

Confidential

DoE IntroductionCore Knowledge(Engineering, Chemistry,…)

Statistical Knowledge

Develop Solutions

DOE is NOT a replacement for process knowledge

Confidential

Questions Answered by DoE

How do we get the best reaction yield ?How much costly catalyst or reagent do we really need ?Can we minimize formation of an impurity?Which experimental factors are relevant?How robust is my process or assay ?

Confidential

DOE: ConsiderationsCan’t replace screening of discreet variables (catalyst, solvent)

Best suited for continuous variableso time, temp, stoichiometry

Not helpful for non-reproducible reactions

Best suited for ‘low maintenance’ reactionso Temp = RT to 150 oCo All reactants added at once

Confidential

DOE: Experimental Objectives

Screeningo Which factors are most influential ? o What are their appropriate values/ranges ?

Optimizationo Extract information on how factors combine to influence

responseo Identify optimized reaction conditions

Robustnesso To assess if small changes in continuous factors have an effect on outcome

Confidential

DOE: Misconceptions

Requires in-depth statistics knowledgeo User-friendly DOE software does this for you

• MODDE (Umetrics)/Design Expert (Stat-Ease)

Confidential

DOE: Misconceptions

Requires in-depth statistics knowledgeo Experimental design software does this for you

Requires lots of experiments and timeo Perhaps. but will always get better quality informationo Typically 11-27 reactions per designo Automation/technology can help reduce the effort

Confidential

DOE: WorkflowDefine an Objective

What issue(s) do you want to resolve?

Define the Factors Prioritize: known, suspected, possibly, unlikelySet HIGH/LOW values for factors (define design space)

Define the Response(s) – how to measure ?Select Experimental Design Generate WorksheetRun the ReactionsPerform Analysis with DOE softwareRun Confirming reaction

Confidential

DOE Design: 24 Full factorial Factor 1 Factor 2 Factor 3 Factor 4

Std A:water B:temp C:time D:acid

% C h equiv

1 0 55 2 3

2 10 55 2 3

3 0 75 2 3

4 10 75 2 3

5 0 55 8 3

6 10 55 8 3

7 0 75 8 3

8 10 75 8 3

9 0 55 2 15

10 10 55 2 15

11 0 75 2 15

12 10 75 2 15

13 0 55 8 15

14 10 55 8 15

15 0 75 8 15

16 10 75 8 15

17 5 65 5 9

18 5 65 5 9

19 5 65 5 9

Confidential

DOE Design: 24 Full FactorialFactor 1 Factor 2 Factor 3 Factor 4

Std A:water B:temp C:time D:acid

% C h equiv

1 0 55 2 3

2 10 55 2 3

3 0 75 2 3

4 10 75 2 3

5 0 55 8 3

6 10 55 8 3

7 0 75 8 3

8 10 75 8 3

9 0 55 2 15

10 10 55 2 15

11 0 75 2 15

12 10 75 2 15

13 0 55 8 15

14 10 55 8 15

15 0 75 8 15

16 10 75 8 15

17 5 65 5 918 5 65 5 919 5 65 5 9

3 Center points

Confidential

DOE Design: 24 Full FactorialFactor 1 Factor 2 Factor 3 Factor 4 Response 1

Std A:water B:temp C:time D:acid yield

% C h equiv %

1 0 55 2 3 44.6

2 10 55 2 3 16.5

3 0 75 2 3 66.9

4 10 75 2 3 65.7

5 0 55 8 3 73.5

6 10 55 8 3 38.9

7 0 75 8 3 64.7

8 10 75 8 3 66.6

9 0 55 2 15 54.3

10 10 55 2 15 82.6

11 0 75 2 15 11.1

12 10 75 2 15 75.9

13 0 55 8 15 1

14 10 55 8 15 80.6

15 0 75 8 15 1

16 10 75 8 15 77.8

17 5 65 5 9 92

18 5 65 5 9 92

19 5 65 5 9 89

Confidential

DOE Creates a Design Space

Confidential

DOE Expts: How Many ?rxns

factors

HI Med Lo Total expts

3 4 3 4 114 8 3 8 195 16 3 16 353 5 7 5 174 10 7 10 275 9 11 9 29

screeningoptim

aztion

Confidential

DOE Publications Trend

22

Total # of OPRD DOE Publications

Confidential

Why is DOE Usage Trending Upwards?

1. Changing R&D Landscapeo Need to do more, with less: efficiency is paramounto Shorter timelines

2. Technology advances enable parallel experimentationo Easier to set up 24 reactions today than 10 years ago

3. Quality by Design (QbD) Movement is here to stayo DOE can create the requisite design space

4. Rising popularity of Green Chemistry Initiatives

23

Confidential

DOE Case Studies

Confidential

Case Study #1: MK-518

First-in-Class Oral HIV-1 Integrase InhibitorApproved by FDA October-12-2007 $ 1.1 B sales in first two years

Confidential

MK-518

Confidential

MK-518

N

N

OOH

O

HN

F

HN

O

NN

O

N

N

OOH

O

H2N OMe

H2N

F

O

NN

OCl

13 step routeChallenge: to reduce manufacture cost by 20%

Confidential

MK-518: Problem step

18 solvents, 8 bases screened (Discreet variables)

Existing Conditions: 4 eq Mg(OMe)2/ 4 eq MeI @ 0.5 M (68% isolated yield)

78 22

Peter Maligres

Confidential

MK-518: DOE Screening

DOE Screening Design Factors: Mg(OMe)2 equiv: 1.0 and 3.0MeI equiv: 2.5 and 5.0Conc: 0.25 and 1.0 MTemperature: 30 and 65 oC

19 reactions

Responses (4 and 20 h):ConversionSelectivity (N vs O)

Confidential

MK-518: DOE Screening

DOE Preferred SettingsBase equiv: 1.0 and 3.0MeI equiv: 2.5 and 5.0Temperature: 30 and 65 oCConc: 0.25 and 1.0 MTime: 4 and 20 h

All factors were relevant

HN

NCbzN

OHO

H MeN

NCbzN

OHO

H N

NCbzN

OHOMe

H

HN

O

F

O

HN

O

HN

F F

DMSO

Confidential

MK-518: DOE Success

Preferred DoE SettingsBase equiv: 1.0 and 3.0MeI equiv: 2.5 and 5.0Temperature: 30 and 65 oCConc: 0.25 and 1.0 MTime: 4 and 20 h

99 1

HN

NCbzN

OHO

H MeN

NCbzN

OHO

H N

NCbzN

OHOMe

H

HN

O

F

O

HN

O

HN

F F

DMSO

20 h

> 90% Assay yield

Confidential

MK 518: Effect of Conc and MeI

Confidential

Effect of Temp & Conc

Confidential

Effect of Base and Conc

Confidential

MK-518: Time Cycle

4 hConv 95%

N vs O 4/1

Confidential

MK-518: In Situ Demethylation

4 h 20 hConv 95% 99%

N vs O 4/1 >200/1

Confidential

MK-518 Concerns

Issues:1. at this higher concentration, end of reaction difficult to stir2. Mg(OMe)2- long term issues with supply & cost3. MeI is mutagenic/carcinogen/toxic

99 1

Confidential

MK-518:Commerical Process

Peter Maligres

Yield = 90%Selectivity = >99.9 %Safer, more economical reagentsIncorporated best practices from DOE:

HI Temp/HI Concentration/Longer reaction times

Confidential

MK-518 Summary

78 22

> 99 < 1

DOE>20% reduction in drug inventory cost achievedHigher Yield cascades back to allow fewer RM/solvents to be usedRunner-up: 2008 Presidential Green Chemistry Award

HN

NCbzN

OHO

H MeN

NCbzN

OHO

H N

NCbzN

OHOMe

H

HN

O

F

O

HN

O

HN

F F

> 90% Yield

Humphrey, Pye et al. OPRD, 2010, in press

Confidential

Case Study #2: NNRTI

Jeff Kuethe

Double deprotection in one-pot

Last step of process; modest, variable yield

Goal: Improve yield; single acid

Approach: HTS, followed by DoE

Confidential

HTS Summary @ 70 oC

Initial screen: 17 acids @ 2 levels/7 co-acids/2 solvents = 180 rxns

Follow-up # 1: 6 acids/3 co-acids = 18 reactions

Follow-up # 2: H2SO4 (3 and 7 eq) in 22 solvents

H2SO4 in MeCN

Confidential

Case Study: NNRTI

Sequential removal gave higher yieldBy-product rejection was important

N NN

NH

O

O

Cl

Cl

NC

N NN

NH

O

O

Cl

Cl

NCH2SO4 (2.2 eq)

MeCN/20 oC95% Yield

O (H2SO4)2

Confidential

Case Study: NNRTI

DOE to optimize 2nd reaction conditions

Confidential

Case Study: NNRTI

Considerations:

1. Water accelerates rate, but too much water precipitates impurities

2. More acid accelerates rate, but too much acid generates impurities

Confidential

Case Study: NNRTI DoE

N NN

NH

O

O

Cl

Cl

NC

N NN

H2N

O

O

Cl

Cl

NC

H2SO4

MeCN

Factors Settings% water 0-10 vol%Temp 55-75 ºCTime 2-8 hAcid 3-15 eq

Response: Assay Yield Screening design (24 full factorial) = 19 reactions

Chosenbased onobservations

Confidential

DOE Design (N=19)

Factor 1 Factor 2 Factor 3 Factor 4 Response 1


% C h equiv %

1 0 55 2 3 44.6

2 10 55 2 3 16.5

3 0 75 2 3 66.9

4 10 75 2 3 65.7

5 0 55 8 3 73.5

6 10 55 8 3 38.9

7 0 75 8 3 64.7

8 10 75 8 3 66.6

9 0 55 2 15 54.3

10 10 55 2 15 82.6

11 0 75 2 15 11.1

12 10 75 2 15 75.9

13 0 55 8 15 1

14 10 55 8 15 80.6

15 0 75 8 15 1

16 10 75 8 15 77.8

17 5 65 5 9 9218 5 65 5 9 9219 5 65 5 9 89

Confidential

DOE Design (N=19)

Factor 1 Factor 2 Factor 3 Factor 4 Response 1


% C h equiv %

1 0 55 2 3 44.6

2 10 55 2 3 16.5

3 0 75 2 3 66.9

4 10 75 2 3 65.7

5 0 55 8 3 73.5

6 10 55 8 3 38.9

7 0 75 8 3 64.7

8 10 75 8 3 66.6

9 0 55 2 15 54.3

10 10 55 2 15 82.6

11 0 75 2 15 11.1

12 10 75 2 15 75.9

13 0 55 8 15 1

14 10 55 8 15 80.6

15 0 75 8 15 1

16 10 75 8 15 77.8

17 5 65 5 9 9218 5 65 5 9 9219 5 65 5 9 89

Confidential 48

Case Study: NNRTI

C (time) has no bearing on outcome

Confidential

Case Study: NNRTIDesign-Expert® SoftwareFactor Coding: Actualyield

Design Points

X1 = A: waterX2 = D: acid

Actual FactorsB: temp = 65.00C: time = 5.00

D- 3.000D+ 15.000

D: acid

0.00 2.00 4.00 6.00 8.00 10.00

Interaction

A: water

yiel

d

0

20

40

60

80

100

120

More water is better than less water

Confidential

Case Study: NNRTI

Factors Settings% water 2-7 %Acid 5-9 eq

11 reactions-Optimiztion designResponses: Conversion and Assay Yield

IncludesCenter point settings

Confidential

Case Study: Conversion

Factor 1 Factor 2 Response 2Std Run A:water B:acid conversion

% equiv %1 2 2 5 992 9 7 5 953 7 2 9 1004 6 7 9 1005 11 2 7 1006 10 7 7 967 3 4.5 5 988 4 4.5 9 1009 5 4.5 7 10010 1 4.5 7 9911 8 4.5 7 95

No relevant model terms

∴Conversion is robust over design space

Confidential

Case Study: Yield

Only water was relevant

Selected:4% water7 equiv acid

Confidential

Case Study: Confirmation

Assay yield = 92% (DoE Prediction: 88.5%)Isolated yield = 90%50 g scale

Kuethe, Weissman et. al. OPRD, 2009, 471.

Confidential

Case Study #2: NNRTI Summary

Acid YieldOriginal 30 eq 65%Improved 9.2 eq 87.5%

70% reduction in acid eq used>35% increase productivity

Confidential

Case Study #3- Suzuki

Goal:1. to reduce Pd(OAc)2 & ligand charges (0.4 mole%,0.8 mole%)2. Improve yield and/or purity

Confidential

Case Study- Suzuki

DOE Factors (Optimization design):Ligand/Pd ratio: 1.0 and 3.0Catalyst load: 0.1 and 0.5 mole%Molarity boronic acid: 0.5 and 1.5Temperature: 60 and 80 oC

27 Reactions in 96-well plate format, 2 days to plan/setup/execute/assay0.65 g material (24 mg/rxn)

Br RR

(HO)2B

Tol/ THF, K2CO3

Pd(OAc)2/Ar-PCy2

70 oC

Confidential

High Throughput Optimization

= 96 x

Confidential

HTS Reaction Vials

Confidential

Case Study- Suzuki

DOE Optimal Settings:Ligand/Pd ratio: 1.0 and 3.0Catalyst load: 0.1 and 0.5 mole% - small effectMolarity boronic acid : 0.5 and 1.5Temperature: 60 and 80 oC (65 oC)

Br RR

(HO)2B

Tol/ THF, K2CO3

Pd(OAc)2/Ar-PCy2

70 oC

Confidential

Effect of Temp and Pd Loading

Lig/ catalyst ratio fixed at 3:1; Triol M fixed at 1.5 M

Overall LCAP

Confidential

Optimized Conditions

Optimized Experiment:-increased LCAP by 1%-decreased Pd by 75%-decreased Lig by 70%

Spencer Dreher

BrR

R

(HO)2B

Tol/ THF, K2CO3

0.1 %Pd(OAc)2

95% AY93 A% pure

65 oC0.25 % Ar-PCy2

1.5 M

Confidential

Case Study #4-Sonogashira

S. Krska/A. Northrup

Medicinal Chemistry conditionsGoal: improve yield via HTS and DoE

Confidential

HTS Result

NH

O

I

NH

O

3

40 mol% CuI3 equiv iPr2NH

MeCN45 oC/18 h

90 A%

10 mol% [(allyl)PdCl]240 mol% (2-furyl)3P

+

10 A%

bis additionimpurity

Screened two discreet factors: ligand and Pd source32 reactions (HTS-96 well plate format) 1.5 days- 125 mg of substrate

Confidential

DOE Optimization

Fixed factors: P/Pd ratio (2:1); temperature (45 oC); base equiv (3)DoE Factors: Optimization Design (17 reactions)• Pd loading (2 and 10 mole%)• Cu/Pd ratio (0.5 and 2.0)

• Alkyne equiv (1 and 5)

Confidential

DOE Optimization

Fixed factors: P/Pd ratio (2:1); temperature (45 oC); base equiv (3)DoE Factors: Optimization Design (17 reactions)• Pd loading (2, 10 mole%)- little effect- set to 3 mole % Pd dimer (6% Pd)• Cu/Pd ratio (0.5, 2.0)- most important; therefore 12 mole% CuI

• Alkyne equiv (1, 5) –2 equiv; less is more

Confidential

DOE Optimization

Cu/Pd = 2

Confidential

DOE Confirming Reaction

DOE Improvements over HTS Result:• 70% reduction in Pd charge• 70% reduction in phosphine charge• 63% reduction in Cu charge• 94% reduction in time cycle (18 h in HTS)• Improved selectivity from 9:1 to 100:1

Confidential

Case Study # 5: Hydrogenation

10% loading Pd(OH)2

25 oC/45 psi/EtOAc (8 vol)

88 A%

Goal: to minimize formation of impurities/maximize desired product

12 A%

Confidential

Case Study # 5

Mark Weisel

88 A%

12 A%DOE Screening design: 4 Factors (19 reactions)1. Temp (25 and 55 oC)2. Pressure (30 and 60 psi)3. Pd(OH)2 loading (5 and 15 wt%)4. Volume EtOAc (6 and 10 ml/g)

Confidential

Relevant Factors: Pd andTemp

-6

-4

-2

0

2

4

6

8

10

12

14

16

18Te

mp Pd Vol

Tem

p*Pd

Pd*V

ol

A%

MODDE 8 - 1/7/2008 4:30:11 PM

Confidential

Effect of Pd, Temp on A% Desired Product

Confidential

Optimal Settings: Confirmation

Cl

CO2H

OMe

Cl

CO2H

OMe

CO2H

OMe

Cl

CO2H

Cl

CO2H

CHO

+ + +H2

Pd(OH)2EtOAc

Mark Weisel

Factors-ranked vs Original Conditions 1. Pd loading (15 wt%) vs 10% 2. Temp (25 oC) vs 25 oC3. Concentration – little effect (10 volumes)4. Pressure –no effect; therefore lowered from 45 to 30 psi

Confidential

Optimal Settings: Confirmation

Cl

CO2H

OMe

Cl

CO2H

OMe

H2 (30 psi)

Pd(OH)2 (15 wt%)EtOAc (10 volumes)25 oC

Mark Weisel

Selectivity improved from 88 A% to >99 A%

Confidential

DOE Benefits

• Increase your process knowledge; more complete picture

• Discover the effects of changing factors

• Understand the relevant interactions

• Can facilitate route selection; define upside

• Save time, materials, and money

Confidential

Take Home Message

DOE is a powerful, efficient approach for optimization of continuous variables at all stages of process developmentThe decision to implement DoE depends on effort required and payoff

High barrier requires big payoff : commercial API processLow barrier requires smaller payoff: Med Chem program

Technology can minimize the effort of running, assaying multiple reactions

Parallel reactors, liquid dispensers, solids dispensing robotics, etc.

Confidential 76

QUESTIONS ?

[email protected]

Design of Experiments (DoE): A “New” Approach to Reaction ... · Design of Experiments (DoE): A “New” Approach to Reaction Optimization Steven A. Weissman NEACS Symposium:

Documents