Automating Chemical Synthesis Through Flow ChemistryNeal SachSenior Principal ScientistOncology Medicinal Chemistry10777 Science Centre DriveLa Jolla, CA 92130, USAneal.sach@pfizer.com

1st RSC/SCI Symposium on Continuous Processing and Flow Chemistry GSK Stevenage, UKThursday 4th November 2010

Contents

Conjure Flow Reactor

Applications and Examples of Conjure Flow Reactor to Discovery and Development

Towards Automated Synthesis

Flow - Key Applications to Drug Discovery

Substrates not accessible using traditional chemical

synthesis techniques

Stab

ility

Reactivity

Medicinal - Expanding Chemical Space Forbidden chemistries Library synthesis Extending temperature ranges (vs. microwave)

Process - Enabling and Economics Scaling Forbidden chemistries (diazotation etc.) Scale-up of Discovery microwave chemistry Minimised scale-up considerations

Early implementation of flow chemistry can significantly ease the transitions from Discovery to Process

Flow Chemistry Limitations

Emerging Technology (at laboratory scale) Heterogeneous ReactionsMaterial RequirementsTime Requirements

2004 Collaboration started with Wyeth and Accendo

nano micro meso kilo lab pilot plant manufacturing

Analytical Scale=0.5ml/min (HPLC like flow rates)

Feed (working solvent)Fluid Prep

interface

Reactortemperature

Discrete segments inserted and maintained in main flow

mix

DiscoveryLibrary synthesis (vary monomers)

ProcessOptimization (vary conditions)

Note mixing is uniquely independent of flow rate

Draw from up to 4 of 36 source vials

Analytics and sample collection

Schematic

Pass? Fail?

One-size-fits-all initial conditions

HPLC/MSProduct Y/N?Yield & Purity

Greater percentage of successful library elements (discovery). Optimized reactions (process)

L H H

DOEcalculates best conditions

L

Library elements (Discovery)Reaction Conditions (Process)

m

n

Vary prescribed reaction parameters around initial conditions. Automatically done based on Simplex and DOE statistical optimization tools.

(flowing segments)

Collect via Fraction Collector (Discovery)Prep via repetitive segments (Process)

Adding Artificial Intelligence

l

Fluid prep module

Reactor module

HPLC stack

Feed module

Conjure Flow Path

Fractioncollector

ELSD

MS

Output flow from Reactor

Sample Dilutor

Sample Dilutor Detail

Total Segment Size 150-800uL. Typically 300uL, 5mg Substrate

Multiple Segments Active Within Reactor (up to 4)

No contamination between reaction segments, key.

Conjure Square-Wave Reaction Segments

50uL He or

FL

300uL ReactionMixture

50uLHe or

FLCarrier Solvent Carrier Solvent

min15 20 25 30 35

mAU

0

200

400

600

800

1000

1200

1400

DAD1 E, Sig=310,2 Ref=off (N70807A\6X6X3TEMPACC 2007-08-07 17-49-10\001-0101.D)

16.43

7

25.6

49

34.49

2

No Dead Reckoning

Peaks On ScaleBubble or

UV Detection

1 2 3 4

Each peak is a 5mg reactionExact Stoichiometry (fluid prep)Exact Reaction Time (flow rate)Exact Temperature (within 0.1C)

Heart cut is sampled and directed for LC/MS Analysis (ultra-fast gradients <2min run-times). Enables real-time results between segments.

Dilutionsyringe2500 L

Samplesyringe25 L

Solvent

Waste

Mixer tube

Primary flow from reactor

Collection or waste

HPLC flow

HPLC column

Sample loop

Injection loop

Sample Dilutor Module

20 L

5 L

Sampling “Heart Cuts” from the Flowing Segments

Mixing with Diluent

Dilutionsyringe2500 L

Samplesyringe25 L

Solvent

Waste

Mixer tube

Primary flow from reactor

Collection or waste

HPLC flow

HPLC column

Sample loop

Injection loop

Sampling “Heart Cuts” from the Flowing Segments

Injecting onto HPLC

Primary flow

Primary flow

Dilutionsyringe2500 L

Samplesyringe25 L

Solvent

HPLC flow

HPL°C flowWaste

Mixer tube

Sample loop

Injection loop

Sampling “Heart Cuts” from the Flowing Segments

Return

Demonstration SNAr Library Preparation

6 Amines x 6 Aryls = 36 Library 0.5M Solutions 100uL Each Total Segment Volume = 300uL 3 Temperatures 150°C, 200°C, 250°C 10 Minute Reaction Time NMP Carrier Solvent 108 Experiments = 15 hours Run On-line HPLC/UV/MS/ELSD Analysis

X Cl

R1R4R2

R3 X N

R1R2

R3

R4

X1

X2

X1 NHX2

NMP

X=C-NO2 or N+ HCl

Hamper, Bruce C.; Tesfu, Eden. Direct uncatalyzed amination of 2-chloropyridine using a flow reactor. Synlett (2007), (14), 2257-2261

150-250°C in-situ Yield by 254nm HPLC/MS Analysis

150°C - 42% 200°C - 61% 250°C - 22%

Overall Library Success Rate : 88%

Demonstration Singleton DOE Optimisation for Process

3 Level 2 Factor Design Temp – 100°C, 150°C, 200 °C Equivalents – 1eq. 2eq. 3eq.

11 Reactions 13-20mg each 2 Centre Points

Total = 184mg Total Time = 116 minutes

N

ClCl

Cl

NH

N

O O

N

N

O O

N

ClCl

+ HCl

NMP

min0 20 40 60 80 100

mAU

0

500

1000

1500

2000

2500

DAD1 A, Sig=254,2 Ref=off (N70809B\DOEACC 2007-08-09 18-43-11\001-0101.D)

1-Vial 8

1-Vial 9

1-Vial 1

0

1-Vial 1

1

1-Vial 1

2

1-Vial 1

3

1-Vial 1

4

1-Vial 1

5

1-Vial 1

6

1-Vial 1

7

1-Vial 1

8

11.55

3

17.98

5

24.25

5

48.76

9

55.19

2

61.59

5 67.84

2

74.17

1

98.14

4

104.7

61

112.0

50

100°C 150°C 200°C

50°C5min

50°C5min

Demonstration Singleton DOE Optimisation for Process

Yield

Investigation: 3,4,5-Trichloropyridine Optimisation (MLR)Contour Plot

66.6320063.25220051.18120035.07215035.41215040.73315035.38215025.2911506.4231005.3721003.411100YieldEquivalentsTemp

0

10

20

30

40

50

60

70

0 10 20 30 40 50 60 70

Obs

erve

d

Predicted

Yield with Experiment Number labels

N=11 R2=0.996 R2 Adj.=0.993DF=5 Q2=0.961 RSD=1.9017

1

2

3

4

5

6

7

8

9

1011

y=1*x+7.531e-006R2=0.9963

Investigation: 3,4,5-Trichloropyridine Optimisation (MLR)

MODDE 7 - 8/14/2007 10:25:40 AM

Predicted

Obs

erve

d

Knife edge optimal conditions 275°C, 2000psi, 15 minutes >50% product

Conclusions Key template accessed IP free chemical space Substrate of interest to

multiple TA’s Flow processes are scalable

Flow De-Carboxylation

Issues Extreme temperatures Extreme pressures Not possible in batch

NN

OH

OH

O

NN

OH

IPA

-CO2

p

21.9291

30.1321

38.3351

38.3351

46.5382

46.5382

54.7412

54.7412

4Sharp Decreasein Yield

Sharp Decreasein Yield

Flow Experimental Accendo DOE optimization

completed in 1hr CCC Design – 15 experiments

300°C200°C Temp5 min

15 minTime

Triazoles Metabolically stable pieces Significantly lower logD of a potential drug

molecule (improved ADME properties) 1,4-substituted-1,2,3-triazoles can be

easily prepared in excellent yield using Click Chemistry

R2 N3

R3

R1

N

NN

R2

R1

R3CuSO4(0.05eq.)NaAscorbate (0.1eq.)tBuOH / H2O

Triazoles and Click Chemistry

Click Chemistry Advantages Extremely fast, clean reactions Installing acetylenes in drug

molecules is convergent with existing and common halide intermediates

Functional group tolerance enables diverse pieces to be stitched together

Disadvantages Preparing low MW organic azides

is extremely dangerous in batch

Kolb, Hartmuth C.; Finn, M. G.; Sharpless, K. Barry. Click chemistry: diverse chemical function from a few good reactions. Angewandte Chemie, International Edition (2001), 40(11), 2004-2021

Azide Formation Can we prepare low MW azides in-situ

from halides and NaN3 in flow ?

The click reaction is known to work with extremely low concentrations of Cu

Triazole Click Chemistry Conditions

DMF / Water required for NaN3solubility and solubility of triazole products

2hr Accendo DOE OptimizationNaXR2 X

R3

R1 N

NN

R2

R1

R3

X=Cl, Br or I

+

NaN3 (1eq.)

5min175°C>75% y

NaXOHBr

N

NN

OH

+

0.25M NaN3

(1eq.)DMF / H2O 10:1150oC / 5min

Scope of in situ Click Chemistry

ICl

BrOH

NH

OCl

N Cl

Cl

N

NO

O

O

O

O

Br

Alkyl Halides

Terminal Acetylenes

0.25M NaN3 (1eq.)DMF / H2O 10:1150°C / 5min

FlowCu Reactor

Bogdan, Andrew R.; Sach, Neal W. The use of copper flow reactor technology for the continuous synthesis of 1,4-disubstituted 1,2,3-triazoles. Advanced Synthesis & Catalysis (2009), 351(6), 849-854

N

NNN

NCl

12, 46 %

N

NNN NH

O

11, 88 %

N

NNN

O

10, 72 %

N

NNN OH

9, 60 %

N

NNN

8, 40 %

N

NNN

7, 70 %

NNN

1, 72 %

NNN

2, 28 %

NNN OH

3, 72 %

NNN

O

4, 43%

NNN NH

O

5, 61%

NNN

NCl

6, 50%

NNN

13, 30 %

NNN OH

14, 76 %

NNN

O

15, 39 %

NNN NH

O

16, 63 %

NNN

NCl17, 60 %

O

O NNN

NCl

23, 77 %

O

O NNN NH

O

22, 63 %

O

O NNN

O

21, 26 %

O

O NNN OH

20, 92 %

O

O NNN

19, 72 %

O

O NNN

18, 51 %

O

ON

NN

24, 50 %

O

ON

NN OH

25, 32 %

O

ON

NNO

26, 21%

O

ON

NN NHO

27, 47 %

N

NNN

NCl

30, 50 %

N

NNN NH

O

29, 65 %

N

NNN

O

28, 58 %

Butyne Sonogashira

Issues Butyne b.p. 8°C Pressurized system Dangerous in batch Selectivity challanges

N

N

I

Br

N

NBrPd(P(Ph)3)4

DIPEAEtOH/Dioxane

"Cu"

+ Cu Reactor 125C, 4 min.

Hastalloy Reactor 125C, 4 min.

Conclusions Scalable process (24g/day) Cu reactor removes

requirements for Cu additive 1.2g delivered to project team

SM

SM

Product

Product

Flow Experimental Accendo optimization

completed in 2hrs Demonstrates 125°C as

optimal for conversion/time

Production Mode

Reproduce automatically optimized chemistry Scale out not up, as a function of time

10 min 500mg 1hour 3g 1day 72g 1week 0.5kg

4 0 6 0 8 0 1 0 0

m A U

0

1 0 0

2 0 0

3 0 0

4 0 0

5 0 0

6 0 0

D A D 1 E , S ig = 3 1 0 ,2 R e f= o f f ( A : \F L O W _ C H . . .O W _ C H E M I S T R Y _ 2 4 6 C L _ D O E \R E P E A T S A C C 2 0 0 7 -0 8 - 1 0 1 5 -1 9 -4 2 \0 0 1 -0 1 0 1 .D )

27.98

6

33.33

6

38.6

98

44.0

75

49.4

12

54.7

88

60.14

5

65.5

66

70.88

2

76.26

3

81.59

8

86.9

78

92.3

41 97.7

45

103.1

51

108.5

35

113.8

82

DOE Software

The Next Step – Automated DOE Optimization

InstrumentControl

Design Wizard

ReagentsDatabase

HPLC/MSAnalysis eLNB

Data Collation

and Interpretation

Closing this loop enables automated

reaction optimization

The Next Step – Automated DOE Optimization

Automated Optimization Set initial DOE space Set theoretical parameter space Set objective (LC/MS)

Enable Automated Simplex Optimization Iteration 1 Iteration 2 Iteration 3

Objective Met System automatically scales

chemistry using optimized conditions

L H

L H

L H

Conclusions

Demonstrated Preparation of discreet square wave reaction segments Library optimization and enumeration (Discovery) Singleton optimization (Process) Singleton scale-out (Discovery and Process) Comment - We don’t do flow, for flows sake

Next steps Continue to identify forgotten chemistries (long forbidden) Close DOE loop for automated optimization Multiple step synthesis

Acknowledgements

Technology Joel Hawkins Jan Hughes (Accendo) Terry Long (Accendo) Kristin Price Larry Truesdale

Chemistry Andrew Bogdan (Cornell) Valery Folkin (Scripps) Jason Hein (Scripps) Peter Huang Kevin Bunker Paul Richardson