Presentation on theme: "Combined use of Design of Experiment (DoE) and Process Automation for the Efficient Optimization of New Synthetic Transformations Universita’ dell’Insubria-Dipartimento."— Presentation transcript:
Combined use of Design of Experiment (DoE) and Process Automation for the Efficient Optimization of New Synthetic Transformations Universita’ dell’Insubria-Dipartimento di Chimica Via Valleggio n o 11 22100 Como (Italy) www.uninsubria.it R&D Chemistry Research Centre Via Lorenzini n o 8 20139 Milano (Italy) www.boehringer-ingelheim.it Literature meeting May 2 nd 2005 Federica Stazi Ph.D Thesis
Reasons for DoE at the Chemistry Research Centre Boehringer Ingelheim Pharma KG CRC Milan, Italy Boehringer Ingelheim Pharma KG Biberach, Germany CRC Drug Development Drug Discovery Intermediates Building blocks Test compounds … Pre-dev. Candidates Metabolites Process impurities … D D iversity O O riented S S ynthesis T T arget O O riented S S ynthesis
Target Oriented Synthesis (TOS) and DoE DoE-driven search for optimal conditions Target
Diversity-Oriented Synthesis (DOS) and DoE … … Same starting material and rxn conditions different RX Same R’X and rxn conditions different starting material DoE-driven search for optimal conditions
The DoE Concept: Basic Principles Inputs x1x1 x2x2 xpxp Outputs System … z1z1 z2z2 zqzq controllable factors uncontrollable factors y … (starting materials) (products)
OFAT (One Factor at A Time) Approach OFAT results in a set of experiment in which only one factors is varied S M P AB C incomplete picture of the overall process factor interactions are not revealed number of experiments not fixed not possible to perform experiments in parallel -
DoE (Design of Experiment) Approach S M P AB C DOE results in a set of pre-planned experiments in which factors are varied at the same time factor interactions are revealed precise estimation of factors effect 2-level Factorial Design 12 34 5 6 87 experimental matrix mathematical model of the chemical process based on statistical analysis possibility to perform experiment in parallel
Doe Simplified: Practical Tools for Effective Experimentation Mark J. Anderson, Patrick J. Whitcomb Productivity Press, 2000 Design and Optimization in Organic Synthesis R. Carlson Elsevier Science, 1997 Design and Analysis of Experiments, 5th Edition D.C. Montgomery Wiley, 2000 + Chemical Journals Statistical Background and DoE Tools
Statistical Background and DOE Tools: Examples S.V. Ley et al. Organic Process Research Development, 2002, 6, 823 R: Et 4 F Res IV, 8 exps + 2 centres A.A. (equiv) PS-DIEA (equiv) Rnx time (hours) Conc.(volumes) Pre DoE:40% Post DoE: 91% 5 different R groups Yields: 81-96% S.V. Ley et al. Synlett, 2000, 11, 1603 5 F ResIV, 16 exps + 4 centres PS-DCC (equiv) Conc. (volumes) Rnx time (hours) Solvent T1 Solvent T2 Post DoE: 97% 8 different R groups 10 different R’ groups 80 cpds. Hit rate 95% 4 F ResIV, 8 exps + 1 centre PS-DCC (equiv) Conc. (volumes) Amine (equiv)
Advantage Series 2050 (Argonaut) SK233 React Array Workstation (Anachem) Carousel (Radley) ? ? ? ? Statistical background and DoE Tools Design Expert 6.0.4 by Stat-Ease MODDE 7.0.0 by Umetrics
React. rack Reagent Solvent racks UV/Vis Detector PC Reaction Control HPLC control Needle Syringes HPLC Statistical Background and DOE Tools
The Sequential Workflow of DoE 2. Planning the experiment: State experimental objectives Choice of factors, levels and response variable choice of experimental design 3. Performing the experiment 4. Data analysis and modeling 5. Interpretation and confirmations 6. Reiteration 1. Synthetic Problem ?
1. cytochrome P450 2. UDPG transferases Putting the Theory into Practice Step 1. Defining the Synthetic Problem: a Problematic Glucuronidation
Putting the Theory into Practice Step 1. Defining the Synthetic Problem: O-Glucuronidation Background UDPG transferases For a review, see: Stachulski, A. V.; Jenkins, N. J. Nat. Prod. Rep. 1998, 173.
Putting the Theory into Practice Step 1. Defining the Synthetic Problem: O-Glucuronidation Background For a review, see: Stachulski, A. V.; Jenkins, N. J. Nat. Prod. Rep. 1998, 173.
Putting the Theory into Practice Step 1. Defining the Synthetic Problem: A New Strategy Modified Koenigs-Knorr cond.: 25% yield (Ag 2 O, mol sieves, 18 h CH 3 CN + TMEDA 10 eq, R=Piv) Typical Koenigs-Knorr cond.: 3% yield (Ag 2 O, mol sieves, 18 h CH 3 CN, R=Ac or R=Piv)
Step 2. Planning the Experiment Find the best starting point: small-scale parallel reagent screening (10 mg scale). Amine vs. “Ag” pKa : 11.0 9.110.310.4 9.2 HMTTA works best. The silver source does not significantly influence yields. influence of amine complexing ability a amine basicity silver source << < a. Meyerstein and al. J. Am. Chem.Soc. 1995, 117, 8353-8361
Step 2. Planning the Experiment: Statement of the Problem State experimental objectives: which type of design? Process screening Process optimization Process robustness testing which variables are most influential? how variables are relevant? Do small changes in uncontrolled variables influence the response?
Step 2. Planning the Experiment: Selection of Factors Choice of factors and factor levels: use of process knowledge + team work Reagents stoichiometry Vol of solvent presence of base mol. sieves rxn time pre-complex. time Type of “Ag” stirr. speed rxn Temp others... “Ag” Br-sugar HMTTA Ag 2 O Ag 2 CO 3 Define design factors, held constant factors, allowed-to-vary factors Factors can be either quantitative (time, stoichiometry) or qualitative (“Ag” type)
Step 2. Planning the Experiment 7 factors to be investigated in a screening factorial design A complete investigation of 7 factors over 2 levels requires: 2 7 = 128 exps 128 parameters are estimable: 1 constant term, 7 linear terms, 21 2-FI, 35 3-FI, 64 4/7-FI FI relative importance: 2-FI > 3-FI >> 4/7-FI
Step 2. Planning the Experiment: Full vs. Fractional Factorial Designs n o of factors n o of experiments 2 3 45 6 78 9 4 8 16 32 64 128 256 Fractional Factorials exploit the redundancy of Full Factorials to reduce the n o of exps 7 factors can also be studied in only a fraction of the original full factorial design. Full Fractional
Step 2. Planning the Experiment: Final Output of Pre-Experimental Plan 7 factors to be investigated in a 2 7-4 Resolution III design: 8 exps + 3 center points (50mg scale) Experimental matrix: center points for curvature detection for calculation of pure error
Step 3. Performing the Experiment Use randomization to reduce the influence of nuisance factors If possible, operate in parallel since we rely on a previous experimental plan Monitor and record values of uncontrolled factors Perform a scoping study: check -- - vs. +++ and reproducibility.
Step 4. Data Analysis and Modeling: ANOVA Testing (Analysis of Variance) of changing variable Ag 2 CO 3 Br-sugar HMTTA
Step 5. Interpretation and Confirmation After stepwise modifying the insignificant terms we obtain the definitive linear model y = + * A+ * C - *D + * E + Is this linear model adequately modeling the response?
Step 6. Reiteration: Altering Factors Ranges The contour plot directs us outside the investigated region modify factors ranges to explore a better experimental region
Different options when the linear model is not adequate. Many are extensions of the 2-level factorial design 2-level FD CCDCCF3-level FD Factor levels 533 Number of Experiments 14+3 27+3 Geometries of the Explored Space sphericalcubic Characteristics: Box-Behnken 3 12+3 spherical Response Surface Modelling (RSM): an Overview
Optimizing Glucuronidation Yield using CCD: Performing the Experiment factorial axial center 20 exps on (100mg scale)
Optimizing Glucuronidation Yield using CCD: Data Analysis and Model Building Definitive coded model yield = 76.91 - 9.58 A + 0.70 B + 2.57 C- 0.75 A 2 + 1.44 A B + 2.51 A 3 Maximum
Optimizing Glucuronidation Yield using CCD: Empirical Model Interrogation Program optimization tools indicate the best conditions found and the confidence intervals FactorNameLevel Low Level High Level AHMTTA0.700.22.5 BAg 2 CO 3 3.763.35.5 CBr-sugar2.422.02.5 PredictionSE Mean95% CI low95% CI high P yield86.51.3483.7189.33 Qty phenol in situ yield isolated yield 1 gr 86.0 80.6 1 gr 87.2 81.0 3.5 gr 85.7 80.0 Model validation
Optimized conditions: Ag 2 CO 3 3.76 eq Br-sugar 2.4 eq HMTTA 0.7 eq 1h CH 3 CN in situ yield 86.0% isolated yield 80.5% Reagents Screening 10 exp DoE Factorial Screening 11 exp DoE CCD Optimization 20 exp Initial conditions: Ag 2 O 2.7 eq Br-sugar 1 eq mol sieves 18 h CH 3 CN isolated yield 3% Optimizing Glucuronidation Yield Using CCD: Conclusion
Ag + Ag 2 O >> Ag 2 CO 3 Ag 2 CO 3 no Br-Sugar Ag + dissolution / activation Ag + competitive complexation Mechanistic Modelling: the Manifold Actions of HMTTA Ag+ active !
Mechanistic Modelling: the Manifold Actions of HMTTA Positive effects of HMTTA : competitive ligand for SM complexation activator of Ag + Max competitive binding to Ag + Max Ag + activation SM Complexation > Ag + activation F.Stazi, G. Palmisano, M. Turconi, S. Clini, and M. Santagostino, J. Org. Chem, 2004, 69, 1097-1103. The postulated irreversible binding of starting material (SM) to Ag + ions is really operative. The presence of the tetramine additive (HMTTA) influences the complexation equilibria. The relationship between complexation of SM and concentration of HMTTA is non-linear. Excess favours the formation of unwanted side product Base (pKa=9.23, 8.47, 5.36, 1.68) on the Br-sugar (-HBr) Negative effect of HMTTA : Consistent depletion of Br- sugar
Other Applications Pd-Catalysed Cyanation of aryl bromide at room temperature F.Stazi, G.Palmisano, M.Turconi, M.Santagostino Tetrahedron Letters, 46 (2005) 1815-1818. Regioseletive Alkylation of 3,4-dihydroxybenzaldehyde Unpublished Results
Summary and Conclusions A mathematical regression model is generated. This model is empirical and valid only within the studied factor range. A better understanding and control of the process are gained by interacting with the model. Use of non-statistical knowledge of the problem for choosing factors and their levels, interpreting the results... “ Using statistics is no substitute for thinking about the problem.” Design and analysis of Experiments D.C. Montgomery DOE results in a set of experiments in which factors are varied at the same time in an organized and systematic approach
Suggestion If you find DoE applied to boring chemistry problem ….. Using DoE to Spend Less Time in The Traffic Screening Ingredients (for Homemade Bread) Most Efficiently with Two- Level Design of Experiment Applied DoE to Microwave Popcorn and more and more…. By Mark J. Anderson, consultant, Stat-Ease, Inc., Minneapolis, MN
Acknowledgment Prof. Giovanni Palmisano Universita’ dell’Insubria-Dipartimento di Chimica Dr. Marco Santagostino Boehringer-Ingelheim R&D Chemistry Research Centre