The time-to-market constraints associated with modern drug development restricts the methodologies and reactions that can be used to make an API. In addition, there are also health and safety issues coupled with environmental factors to consider. An ideal reaction would use an innocuous catalyst to convert all of the substrate to the desired product. However, we are far from this ideal situation. A catalytic reaction has many advantages compared with one that uses a stoichiometric reagent: one molecule of catalyst can produce many molecules of product. This decreases the amount of reagent needed, which can have positive consequences for waste and cost.

For many transformations, it is no trivial undertaking to find a workable catalyst, let alone an ideal one, and time constraints may not allow a large number of possibilities to be screened. If the step is in the middle of a proposed reaction sequence, the whole sequence could fail if a suitable catalyst is not found. To minimize these problems, we have undertaken a programme that rapidly screens both biological and chemical catalysts.

There are a number of components that need to be put into place to quickly and successfully screen potential catalysts. The first is a large library of catalysts. For biological systems these can be enzymes or organisms, and in some cases organisms that express DNA can be used to make the enzymes without isolation before being tested in the screen.¹ On the chemical side, organocatalysts are becoming increasingly popular, but there are currently relatively few examples. These can be purchased or synthesized and then kept in what is called a "hotel" for screening.

Chemical catalysts based on metals usually have a ligand associated with the metal. This type of catalyst can either be preformed or made in situ. Again, the catalysts or ligands can be kept in a hotel before being used in a screen.² There is another alternative, discussed in this article, where the ligands are prepared and used as part of the total screening process.

One of the major problems associated with the number of possible catalysts is how large this number can be. The number of reactions that have to be performed would take too long to be done by a single laboratory worker in a traditional manner. For example, using five different metal sources with just 10 different ligands produces 50 different combinations. To save time, robots have to be used to run reactions in parallel.

Just performing a large number of reactions at the same time is still not a solution to the problem because the bottleneck is moved to the next step. The reactions have to be analysed to see if the desired product is formed and at an acceptable yield. For biological systems where whole cells are involved, it may be possible to design a screen where the organism will only grow if the desired reaction occurs.

An alternative is to use a colorimetric method where a product or by-product causes a dye to change colour, thus identifying the useful organism. An example of such an approach is the enzymatic hydrolysis of an ester that produces an alcohol and carboxylic acid. The formation of the latter can cause a change in pH, which can be detected by adding a visual pH indicator. For chemocatalysis, using visual methods can be more difficult to implement and chromatographic methods have to be adapted. As the goal of the screening exercise is to find a catalyst that has potential, rigorous analysis is not required.

For gas chromatography (GC) and high performance liquid chromatography (HPLC), using short monolith columns can cut analysis times from tens of minutes to less than 1 min. The time needed to analyse a 96-well plate is, therefore, reduced from days to less than 2 h.

Spectroscopic methods can also detect the presence of the desired product and one that has proven useful is flow nuclear magnetic resonance (NMR) spectroscopy. In this technique, solvent is pumped through tubing in the NMR's probe. A robot then injects the samples from reactions under investigation into this solvent flow for detection. Obviously, there needs to be signals that are readily apparent for the reaction to be detected, either by their appearance in the product or their disappearance.