The document discusses how design of experiments (DOE) can help optimize chemical reactions and processes in a more efficient manner than traditional trial-and-error approaches. It provides examples of how pharmaceutical companies have used DOE to reduce costs and improve yields for key reaction steps. DOE allows researchers to systematically vary multiple reaction factors at once to understand their effects and interactions, requiring fewer total experiments than one-factor-at-a-time testing.

1. Design of Experiments (DOE): A “New” Approach to Reaction Optimization Dr. Steven Weissman Merck & Co. Feb 4, 2008/UPR

2. Outline Background: Big changes for Pharma Basic Principles of Design of Experiments Merck Case Studies Take Home message/Questions

3. Big Changes for Big Pharma Costs/Risks of drug development are rising low hanging fruit has been picked small molecules no longer in vogue protein-based and vaccines = more opportunities ‘ Vioxx hangover’- more trials/more patients = $$

4. Big Changes for Big Pharma Costs/Risks of drug development are rising Globalization of marketplace US market sales is matured; slow growth Emerging markets sales = high growth potential Strategic portions of drug development outsourced to India/China (low-cost providers)

5. Big Changes for Big Pharma Costs/Risks of drug development are rising Globalization of marketplace Uncertain pipelines ‘Batting average’ is unchanged despite huge investments Is bureaucracy killing drug discovery? Are smaller companies becoming better at this?

6. Big Changes for Big Pharma Costs/Risks of drug development are rising Globalization of marketplace Uncertain pipelines Revenues/profits are being squeezed patent expirations – Fosamax ($3 B) tougher regulatory environment payers demand value-added lower cost structures: aggressively pursued downsizing plant closings-namely here in PR

7. Big Changes for Big Pharma Costs/Risks of drug development are rising Globalization of marketplace Uncertain pipelines Revenues/profits are being squeezed

8. New Approaches Needed What can we do as chemists to change the way we do our jobs ? Can we work smarter/faster ? How ??

9. New Approaches Needed What can we do as chemists to change the way we do our jobs ? Can we work smarter/faster ? How ?? Automation/Technology

10. Automated Synthesis Cycle Design Experiment Analysis Informatics

11. Automated Synthesis Cycle Design Experiment Analysis Informatics Design of Experiments

12. Current Approach to Optimization Change One Factor at a time (OFAT) Rarely leads to optimal conditions Leads to different conclusions depending on starting point Requires many expts/little information Cannot separate “noise” from true variability

13. Current Approach to Optimization Change One Factor at a time (OFAT) Rarely leads to optimal conditions Leads to different conclusions depending on starting point Requires many expts/little information Cannot separate “noise” from true variability Ignores interactions of variables

14. Example of OFAT (11/07)

15. 21 Reactions

16. DOE vs OFAT OFAT: 3 factors needed 21 reactions No information on interactions of effects DOE: 3 or 4 factors- 11 or 17 reactions Better quality information Learn about interactions of effects Fewer reactions

17. Notable Quote “ If you test one factor at a time, 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 in 1996

18. What is DOE ? Selected set of expts in which all relevant factors are varied simultaneously ‘ Continuous’ factors are ideal (time, temp, equiv)-the ‘ How much’ Analysis reveals which factors influence the outcome and identifies optimal conditions Systematic, organized approach to problem solving Mathematical model of the design space

19. DoE Introduction Core Knowledge (Engineering, Chemistry, Biology,…) Statistical Knowledge Develop Solutions DOE is NOT a replacement for process knowledge

20. Questions to be Answered by DoE How do we get the best synthetic yield ? How much catalyst/ligand do I need ? Can we minimize formation of an impurity? Which experimental factors are (un) important? How robust is my process ?

21. DOE: Considerations Can’t replace full screening of catalyst or solvent (HTS)- ‘ discreet ’ variables Best suited for continuous variables time, temp, stoichiometry Not helpful for non-reproducible rxns Best suited for ‘low maintenance’ rxns Temp = 20 to 150 o C All reactants added at once

22. DOE: Experimental Objectives Screening Which factors are most influential ? What are their appropriate values/ranges ? Optimization Extract information regarding how factors combine to influence response Identify optimized reaction conditions

23. DOE: Misconceptions Requires in-depth statistics knowledge User-friendly DOE software does this for you MODDE (Umetrics)/ Design Expert (Stat-ease)

24. DOE: Misconceptions DoE requires in-depth statistics knowledge Experimental design software does this for you DoE requires a lot of experiments and time Perhaps. but will always get better quality information Typically 11-27 reactions per design Automation/technology helps reduce the effort needed

25. High Throughput Screening = 96 x Discreet variables- ‘The what” What is the best ligand/catalyst combination ? What is the best solvent ?

26. High Throughput Screening = 96 x Can we do OPTIMIZATION this way too ??

27. High Throughput Optimization ?? = 96 x If so,………………….. Which reactions do we run ? How do assess the data ?

28. High Throughput Optimization ?? = 96 x Statistical Design of Experiments (DOE)

29. HTS Reaction Vials

30. DOE: Workflow Define the Objective screening, optimize, robustness Definition of Factors Prioritize: known, suspected, possibly, unlikely Set HIGH/LOW values for factors (define design space ) Define the Response – how to measure ? Select Experimental Design Generate Worksheet Run the Reactions Perform Analysis with DOE software

31. DOE Design (N=27) 64 0.25 1.85 0.875 4 100 20 58 0.25 1.85 0.875 2 100 19 66 0.25 1.85 0.875 3 115 18 64 0.25 1.85 0.875 3 85 17 76 0.4 2.5 1.5 4 115 16 80 0.1 2.5 1.5 4 85 15 65 0.1 2.5 1.5 2 115 14 79 0.4 2.5 1.5 2 85 13 76 0.1 2.5 0.25 4 115 12 88 0.4 2.5 0.25 4 85 11 63 0.4 2.5 0.25 2 115 10 48 0.1 2.5 0.25 2 85 9 54 0.1 1.2 1.5 4 115 8 39 0.4 1.2 1.5 4 85 7 55 0.4 1.2 1.5 2 115 6 45 0.1 1.2 1.5 2 85 5 55 0.4 1.2 0.25 4 115 4 43 0.1 1.2 0.25 4 85 3 49 0.1 1.2 0.25 2 115 2 45 0.4 1.2 0.25 2 85 1 Yield E:Conc D:Boron/Br C:Cu B:P/Pd A:temp Rxn # Response 1 Factor 5 Factor 4 Factor 3 Factor 2 Factor 1

32. DOE Creates a Design Space Design-Expert® Software Yield X1 = A: temp X2 = B: P/Pd X3 = C: Cu load Actual Factors D: Boron/Br = 2.50 E: Conc = 0.40 Cube Yield A: temp B: P/Pd C: Cu load A-: 85.00 A+: 115.00 B-: 2.00 B+: 4.00 C-: 0.25 C+: 1.50 63.7936 74.1825 86.3492 83.738 59.9047 70.2936 82.4603 79.8492

33. DOE Expts: How Many ? screening optimaztion 29 9 11 9 5 27 10 7 10 4 17 5 7 5 3 35 16 3 16 5 19 8 3 8 4 11 4 3 4 3 factors Total rxns Lo Med HI rxns factors

34. DOE Case Studies

35. MK-0518 First-in-Class Oral HIV-1 Integrase Inhibitor Approved by FDA October-12-2007

36. MK-0518

37. MK-0518 Challenge: to reduce manufacture cost by 20%

38. MK-518: Problem step Peter Maligres Existing Conditions : 4 eq Mg(OMe) 2 / 4 eq MeI @ 0.5 M (68% isolated yield) 18 solvents, 8 bases screened 78 22

39. MK-518: DOE Optimization Peter Maligres DOE Optimzation Design Factors : Mg(OMe) 2 equiv: 1.0 and 3.0 MeI equiv: 2.5 and 5.0 Conc: 0.25 and 1.0 M Temperature: 30 and 65 o C 19 reactions Responses (4 and 20 h): Assay yield Selectivity

40. MK-518 Optimization Peter Maligres DOE Optimal Settings Base equiv: 1.0 and 3.0 MeI equiv: 2.5 and 5.0 Temperature: 30 and 65 o C Conc: 0.25 and 1.0 M Time: 4 and 20 h 99 1

41. MK 518: Surface Model

42. Effect of Temp & Conc

43. Effect of Base and Conc

44. MK518-In Situ Demethylation 99/1 80/20 N vs O 99% 95% Conv 20 h 4 h

45. MK-518 Concerns Peter Maligres Issues: 1. at this higher concentration, end of reaction difficult to stir Mg(OMe) 2 - long term issues with supply & cost MeI is mutagenic/carcinogen/toxic 99 1

46. MK-518 Optimization Peter Maligres Yield =90% Selectivity = >99.9 % Safer, more economical reagents Incorporated best practices from DOE: HI Temp/HI Concentration/Longer reaction times

47. MK-518 Summary 78 22 > 99 < 1 DOE Goal of 20% reduction in drug inventory cost was achieved Higher Yield cascades back to allow fewer RM/solvents to be used Submitted for 2008 Presidential Green Chemistry Award

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

49. Case Study 2- Suzuki DOE Factors : Ligand/Pd ratio: 1.0 and 3.0 Catalyst load: 0.1 and 0.5 mole% Molarity boronic acid: 0.5 and 1.5 Temperature: 60 and 80 o C 27 Reactions in 96-well plate format, 2 days to plan/setup/execute/assay 0.65 g material (24 mg/rxn) !!

50. Case Study 2- Suzuki DOE Optimal Settings : Ligand/Pd ratio: 1.0 and 3.0 Catalyst load: 0.1 and 0.5 mole% Molarity boronic acid : 0.5 and 1.5 Temperature: 60 and 80 o C ( 65 o C )

51. Effect of Temp and Pd Loading Lig/ catalyst ratio fixed at 3:1; Triol M fixed at 1.5 M Overall LCAP

52. Optimized Conditions Optimized Experiment: -increased LCAP by 1% -decreased DesBr impurity (50%) -decreased Pd by 75% -decreased Lig by 70% Spencer Dreher

53. Case Study #3 Dave Pollard Goal: to reduce cost by increasing productivity 100 g/L

54. Screening Design Dave Pollard Factors Octanol: 40 and 60 % NADP equiv: 0.1 and 0.5 % Concentration: 50 and 150 g/L Temp: 25 and 35 o C Enzyme load: 0.3 to 1.0 g/L 19 experiments

55. DOE Factors Plot

56. Interaction: Conc and NAD

57. Screening Result Dave Pollard Factor Preferred Setting Octanol: 40 and 60 % No impact NADP equiv: 0.1 and 0.5 % No impact- increase more ? Concentration: 50 and 150 g/L 50 g/L- undesirable setting Temp: 25 and 35 o C minimal effect- set at 30 o C Enzyme load: 0.3 to 1.0 g/L 1.0 g/L- increase more

58. Optimization Design Dave Pollard Factor NADP equiv: 0.5 and 1.5 % Concentration: 100 and 200 g/L Enzyme load: 0.5 to 3.0 g/L 19 experiments

59. Optimization Design Factor Preferred setting NADP equiv: 0.5 and 1.5 % No effect Concentration: 100 and 200 g/L 200 Enzyme load: 0.5 to 3.0 g/L 3.0

60. Optimization Design Factor Preferred setting NADP equiv: 0.5 and 1.5 % No effect Concentration: 100 and 200 g/L 200 Enzyme load: 0.5 to 3.0 g/L 3.0 Confirming experiment at 200 g/L NADP= 0.5 g/L and enzyme at 3 g/L gave 100% conversion Goal achieved

61. Case Study #4-Sonogashira S. Krska/A. Northrup Medicinal Chemistry conditions

62. HTS Result Screened: ligands and Pd sources 32 reactions (HTS-96 well plate format) – 1.5 days- 125 mg of substrate

63. DOE Optimization 3 Factor RSM design – 19 reactions Fixed factors: L/Pd ratio (2:1); temperature (45 o C); base equiv (3) Varied Factors: Pd loading (2 and 10 mole%) Cu/Pd ratio (0.5 and 2.0) Alkyne equiv (1 and 5)

64. DOE Optimization 3 Factor RSM design – 19 reactions Fixed factors: L/Pd ratio (2:1); temperature (45 o C); base equiv (3) Varied Factors: Pd loading (2 and 10 mole%)- little effect- set to 3 mole% Cu/Pd ratio (0.5 and 2.0 )- most important therefore 12 mole% CuI Alkyne equiv (1 and 5) – 2 equiv

65. DOE Optimization Cu/Pd = 2

66. DOE Confirmation Confirming reaction run using iChem Explorer to monitor reaction DOE/Automation Improvements over HTS Result: 70% reduction in Pd charge 70% reduction in ligand charge 63% reduction in Cu charge 94% reduction in time cycle Improved selectivity from 9:1 to 100:1-mostly due to time aspect

67. iChem Explorer Hardware: heating (to 150 o C) and stirring block for HP 1100/1200 systems Software: to visualize data up to 100- 1 mL reactions in LC vials monitor by direct injection

68. DOE Summary

69. Case Study # 5 Mark Weisel 10% loading Pearlman’s catalyst 25 o C/45 psi/EtOAc 88 A% Goal: to minimize formation of impurities/maximize desired product 12 A%

70. Case Study # 5 Mark Weisel 88 A% 12 A% DOE design: 4 Factors (19 reactions) Temp (25 and 55 o C) Pressure (30 and 60 psi) Pd(OH) 2 loading (5 and 15 wt%) Volume EtOAc (6 and 10 ml/g)

71. Factors

72. Effect of Pd and Temp

73. Optimal Settings Mark Weisel Relevant Factors-ranked Pd loading ( 15 wt% ) Temp ( 25 o C ) Volume ( 10 ml/g ) Pressure- no effect- run at 30 psi Selectivity improved from 88 A% to > 99 A%

74. DOE Benefits Increase your process knowledge Discover the effects of changing factors Understand the effects of interactions Learn what is and what is NOT important Save time, materials, and money

75. Take Home Message DOE is a powerful tool for optimization of reactions Automated tools minimize the effort of running multiple rxns HTS & DOE in 96-well format represents leading-edge science Academics embracing HTS approach Professor D. MacMillan ( Princeton)

76. Acknowledgments Peter Maligres Danny Gauvreau Spenser Dreher Dave Pollard Shane Krska Mark Weisel Dave Tellers

77. QUESTIONS ?