Route Scouting Screening Optimization and Validation

Recently, a large body of work was published describing medium-throughput screening (ten at a time) applied to process development. Not surprisingly, this approach has prompted much effort because a benefit can be obtained from increasing experimental throughput by just one order of magnitude. Therefore, systems are being developed that allow for the set-up, control, and analysis of "ten-at-a-time'' reactor modules. Fortunately, true combinatorial exploration of the very large parameter space is often not necessary for process optimization, therefore this arena provides greatly improved productivity before encountering the real issues of high-speed analytical and data management required to advance to high-throughput screening (100s or 1000s at a time).

Owen and his group at GlaxoSmithKline described a step-by-step approach to the optimization of a synthetic transformation using a central composite experimental design, in conjunction with a standard Gilson 231XL autosampler and automated online HPLC [9]. This procedure aids the appropriate decision-making at each phase from the route scouting to the process validation and finally to the pilot or real plant production. The reactions specified by the experimental design model were prepared by hand. Aliquots of reactions were placed in the Peltier block sample tray of the autosampler, and the progress of each reaction was monitored automatically. The analysis of the data obtained from these semiautomated experiments was performed by commercially available DOE software package Design-Expert 5 (DX-5) (, resulting in highly predictive models for the reaction after just 6 days of experimentation. A set of the preferred conditions was first validated in traditional glassware in the laboratory and then subsequently at the pilot plant.

Fig. 31.1. Illustration of hard- and software package developed by Symyx Technologies for scaled up secondary screening of hits identified in primary screening. Bottom left, array of 96 pressure reactors; top left, table with conversion data for 48 parallel reactions; top right, graphical representation of conversion data from the table.

Fig. 31.1. Illustration of hard- and software package developed by Symyx Technologies for scaled up secondary screening of hits identified in primary screening. Bottom left, array of 96 pressure reactors; top left, table with conversion data for 48 parallel reactions; top right, graphical representation of conversion data from the table.

This study showed that it was possible to carry out reactions in a ''reaction station'' located within the working envelope of an autosampler. Although adequate for this study, the Gilson 231XL and the Peltier rack had severe limitations in the feasibility of other chemistries. The study also showed that having minimized the bottleneck of analyzing the samples, reaction preparation and data manipulation became the labor-intensive task. For all of these reasons, researchers at Glaxo Wellcome designed the development automated reaction toolkit (DART) and the process research optimization screening parallel experimentation robot (PROSPER) systems, which are specifically designed to accelerate process optimization using automation and experimental design.

The DART system became commercially available from Anachem as SK233TM. The SK233™ consists of one or two STEM reaction blocks and a Gilson 233XL Autosampler. Each STEM reaction block allows up to ten reactions at a temperature of —30 °C to 150 °C. The SK233TM was designed to deliver liquids to each reaction and perform online automated HPLC analysis without attendance [16]. Researchers at SmithKline Beecham extended the ability of the SK233TM workstation by improving the performance of the SK233™ under reflux conditions. New glass reaction vessels with a quick-fit female joint and a cold finger-type reflux condenser with a hollow center were implemented in the SK233TM [17]. This multiple condenser arrangement is now known as the REACTarray™ (the REACTarray™ is available under license form Anachem Ltd; The REACTarrayTM allows reactions under reflux conditions and inert atmosphere, as well as reagent additions and sampling under the same conditions.

Researchers at Glaxo Wellcome performed process screening and optimization, robustness tests, reaction profiling, and stability testing using the SK233 in conjunction with statistical experimental design (i.e. DOE). In order to optimize, for example, a Mitsunobu reaction, 20 experiments were run based on a two-level factorial design over 5 days, followed by 16 experiments over 4 days. These results helped to identify an important reaction factor, and the yield was improved from less than 70% to almost 90% [16].

A group at SmithKline Beecham performed reaction scouting, process screening, and process optimization using the REACTarray and the SK233. In one example, the screening of air-sensitive reagents and solvents for a Lewis acid-catalyzed reaction between enol ether 7 and b-lactam 8 was carried out (Scheme 31.2). The screening of 20 Lewis acids and seven solvents provided a number of effective new catalytic systems. During this study, the large amounts of data generated (420 chro-matograms) were handled using Excel [17]. Researchers at SmithKline Beecham described additional applications of multireactor systems in their presentation at the CombiCat conference 1999 [18]. The catalyst, solvent, amine stoichiometry, and catalyst loading were optimized for a sensitive hydrogenation in the synthesis of famciclovir.



Lewis acid solvent

Scheme 31.2. Medium-throughput screening in the synthesis of Elanapril (9) (Merck).

Glaxo Wellcome's new PROSPER system has over 50 reactors that are individually controlled with features to enable rapid process development [19]. One example is the optimization of a phase transfer-catalyzed alkylation to produce a secondary amine.

In the process chemistry group at Eisai, the more than 30-step synthesis of an endotoxin antagonist was carried out by a team of four to six chemists by employing automation-assisted DOE, which can be a very tedious task using conventional manual operations. An array reactor that runs 12 parallel reactions with flexible liquid handling, independent temperature control, timed sequences, and an HPLC interface was used to apply this methodology to optimize carbodiimide-mediated amide bond formation using very precious starting materials; the amine 10 for this reaction was prepared in 12 steps and the acid 11 was purchased at $25,000 per kg (Scheme 31.3). Five factors were targeted for DOE and 20 experiments were run in two batches of ten experiments each. The analysis of data by the statistical software identified the temperature as the only significant factor. The verification runs were also carried out in the array reactor, and the further optimization led to the successful pilot plant synthesis on a 1-kg scale [20].

0O O


Was this article helpful?

0 0

Post a comment