BUILDING the perfect model

Pharmaceuticals manufacturers are increasingly turning to sophisticated modelling techniques to help design their plants and processes, as Stuart Nathan reports.

The process industries are full of apparent paradoxes, but one of the most striking is the complexity of the methods and equipment needed to make the simplest molecules; and the simplicity of those for the most complex substances. The most complicated molecules produced by the industry - the active ingredients of pharmaceuticals - are generally made by batch processes.

Although many aspects of these processes - notably the instrumentation devices used to monitor reactions and the attendant computers which control the process - have advanced considerably, the processes themselves are fundamentally unchanged from the ones used fifty years ago. Often, the processes are literally scaled-up versions of the benchtop experiments used to synthesise the original tiny amounts of the substances for tests.

Pharmaceutical companies are now beginning to rethink this deeply entrenched approach to their processes. The computer modelling techniques, now becoming familiar in the design of large chemical plants, are coming into play in the batch field, and are likely to infiltrate all parts of the pharmaceuticals industry.

In the UK, for example, the Engineering and Physical Sciences Research Council is funding a project known as BRITEST (Batch Route Innovation Technology Evaluation and Selection Techniques), which has the ambitious goal of changing the way that companies design and assemble batch plants.

The BRITEST consortium includes Rhodia, Zeneca, Glaxo Wellcome and Synetix; eight companies which supply intermediates to the industry are also involved, including Bush Boake Allen, FMC Process Additives, Hampshire Chemicals, Pentagon Chemicals and MacFarlane Smith. Academic muscle comes from the business school and Keyworth Institute at Leeds University (the latter is a consortium of departments including chemical and mechanical engineering), the chemical engineering department at UMIST, and the Centre for Process Systems Engineering at Imperial College London.

The best and the BRITEST

BRITEST is a three-year project valued at some £1.3million. The goals are very concrete: to halve the time taken to develop a project, from start of process development to manufacture; to reduce the total manufacturing time for producst, from the time raw materials enter the plant to when the finished products leave the factory; to reduce the capital cost of new plant by 30 to 40 per cent; and to design plants that are 'inherently more versatile', and can be reconfigured for new products and different processes more quickly.

Imperial College is handling the bulk of the modelling work, incorporating process design from UMIST and batch plant designs from Leeds. The intention here is to develop tools that can assess the versatility and reconfigurability of the plants, including risk analyses. The Leeds Business School will also be involved in the latter, evaluating the impact of these radical new designs on the businesses themselves.

Cascade effect

Although most of the industrial partners in BRITEST are involved in the pharmaceuticals industry, the results of the project are likely to benefit the industry as a whole. With companies tending to move away from traditional complex and expensive continuous plants, except for the bulk materials like polymers and major intermediates, techniques borrowed from the drugs manufacturers have become much more widespread. BRITEST could usher in a breath of fresh air, and help ease margins in these congested markets.

Manufacturing isn't the only area where computerised techniques are beginning to hold sway. Active ingredients are only one of the components needed to produce a usable drug, and formulation - the combination of actives, binding agents, excipients and so on - is as precise an art as any other. It's a job that can normally only be handled with the input of an expert.

Some of the requirements of formulators can be mind-boggling. For example, a tablet will need to have a certain strength, so it can survive its packaging process and remain intact during transportation. It must be very stable over a precisely-known shelf-life. It will have to disintegrate at a precise rate inside the body, so that the active ingredient is delivered correctly. Every ingredient of the tablet will affect these properties, and the formulation machinery must be programmed with the right proportions of the ingredients.

Such procedures are increasingly the province of expert systems. In theory, the knowledge gleaned by an expert formulator could be encoded into a computer memory, which would then tell manufacturers the best way to combine their product to achieve the desired results.

This case-based reasoning (CBR) approach is the subject of much research in many different fields. It is part of the array of techniques that come under the umbrella of artificial intelligence, because CBR requires the computer to learn from previous experience. In a nutshell, CBR solves problems by comparing them to similar problems it has solved in the past, and adapting the solutions it used then.

Suitable case for treatment

In a recent article in Pharmaceutical Technology Europe, RC Rowe of Zeneca and S Craw and N Wiratunga of the school of computer sciences at the Robert Gordon University, Aberdeen, evaluated an off-the-shelf CBR package as a tool for formulation. The package, ReCall, was developed by French firm Isoft. Rowe and colleagues tested the program by 'teaching' it a series of formulations of eight doses of twelve different drugs, and asking it to devise formulations for another eight doses of a thirteenth. The results, they say, were surprisingly effective, although errors crept in for high-dose formulations. With a little adaptation, however, the system could handle many formulation tasks, they said. PE