(421c) Intensifying Process Development Using High Throughput Technologies

The objective of High Throughput Technologies (HTT) is to allow a high number of experiments to be completed in the shortest time period possible, achieved by performing experiments in parallel using robotic workstations. HTT has shifted the bottleneck in process R&D away from screening to a problem with high throughput process development. One large pharmaceutical company recently reported(i) that 85% of the successful chemical syntheses emerging from automated discovery are currently rejected for commercial processes. The reasons for rejection are diverse, including economic viability, difficulties in downstream processing, process safety and environmental impact. As the number of compounds screened in the search for new catalysts, chemicals and pharmaceutical actives grows, the pressure to reduce this high rejection rate is increasing. To extend the application of HTT from discovery to product and process R&D is a multifaceted challenge involving many areas core to chemical engineering unit operation design.

In addition to the increase in laboratory throughput, by several orders of magnitude, advances have been made in software tools for molecular modelling in discovery and in the application of simulation technologies based on chemical engineering principles for design, scale up, optimisation and control. The integration of high-speed experimentation with novel experimental design procedures combined with statistical data mining, predictive modelling and simulation methods has the potential of delivering intensified process development (IPD).

Reaction mechanism is a prerequisite for the regression of kinetic parameters and the use of process modelling and simulation tools, providing a link between chemistry and chemical engineering. Although commercial software exists for kinetic fitting, the elucidation of reliable reaction mechanisms from complex networks of reactants and products remains a limiting step requiring considerable time and intellectual effort. This effort may be reduced facilitating IPD with the use of robotic workstations to provide data for automation of reaction mechanism determination.

Finding a useful model of a number of interacting chemical reactions (i.e. a reaction mechanism) is mathematically equivalent to a network inference problem, and may be redefined as the identification of the structure and parameters of a reaction network. Crampin et al. (2004) provide a survey of mathematical and computational techniques proposed to deduce complex biochemical reaction networks. Typically, these methods require the generation and analysis of significant quantities of experimental data in terms of composition-time series (e.g. Arkin and Ross, 1995; Arkin et al, 1997; Samoilov et al. 2001; Tsuchiya and Ross, 2001; Kremling et al., 2004; Bendtsen, et al. 2001). In particular, the methods based on system perturbations around stationary states do not lend themselves to the type of experimentation performed in most robotic workstations, i.e. lab scale batch reactors.

Crampin et al. (2004) suggest a general approach that appears suitable for batch experimentation. They suggest a methodology involving the selection of specific terms from global models comprising sets of basis functions, where each set corresponds directly to the set of all possible reaction rate terms involving the chemical species. Burnham et al. (2006) and Searson et al. (2006) propose and develop a similar approach that uses mathematical and statistical methods to iteratively select the basis functions and identify the corresponding parameters. However, this work did not take into account a number of practical constraints associated with robotic workstations.

The objective of this paper is to present an evaluation of the extent to which HTT experiments may be used to provide the necessary quantitative understanding of both physical and chemical phenomena required for IPD. This will be achieved using a robotic workstation (the Chemspeed SLT 106 synthesizer) and reaction calorimeter with the mechanism elucidation procedure proposed by Burnham et al. (2006) and the BatchCAD modelling and simulation software It is recognised that commercial automated jacketed reactor systems with vessel sizes of 50ml and above provide efficient stirring, accurate dosing and individually instrumented and computer controlled cooling. These are generally considered to provide reliable data for process scale up studies. Although small scale HTT equipment may have this capability, this does not necessarily translate into the capacity (rates of, mixing, mass and heat transfer, accurate dosing, sampling accuracy etc.) necessary to provide reliable data for process scale up. A case study using an L-proline catalysed aldol reaction will be presented to illustrate the work.

References

Arkin, A. and Ross, J. (1995) Statistical construction of chemical reaction mechanisms from measured time-series, J. Phys. Chem., 99, 970-979. Arkin, A., Shen, P. and Ross, J. (1997) A test case of correlation metric construction of a reaction pathway from measurements, Science, 277(5330), 1275-1279. Bendtsen, A.B., Glarborg, P., Damjohansen, K. (2001) Visualization methods in analysis of detailed chemical kinetics modelling, J. Comp. Chem., 25, 161-170 Burnham, S.C., Searson, D.P., Willis, M.J. and Wright, A.R. (2006) Towards the automated elucidation of chemical reaction mechanism from batch reactor experiments, AIChE Process Development Symposium, Palm Springs. Crampin, E.J., S. Schnell, S. and McSharry, P.E. (2004) Mathematical and computational techniques to deduce complex biochemical reaction mechanisms, Progress in Biophysics & Molecular Biology, 86, 77?112. Kremling, A., Fischer, S., Gadkar, K., Doyle, F., Sauter, T., Bullinger, E., Allgöwer, F. and Gilles, E.D. (2004) A benchmark for methods in reverse engineering and model discrimination: problem formulation and solutions, Genome Res., 14, 9, 1773-1785. Samoilov, M., Arkin, A., Ross, J. (2001) On the deduction of chemical reaction pathways from measurements of time series of concentrations, J. Chaos, 11, No. 1, 108-114. Searson, D.P., Burnham, S.C., Willis, M.J. and Wright, A.R. (2006) Identification of chemical reaction mechanism from batch process data, The 17th IASTED International Conference on modelling and simulation, Montreal, May. Tsuchiya, M., Ross, J. (2001) Application of genetic algorithm to chemical kinetics, J. Phys. Chem., 105, 4052-4058

(i)Private communication, Allen Wright, 2004