Abstract
INTRODUCTION
The throughput of many of the functional elements at the heart of the drug discovery process, including compound storage and supply, combinatorial chemistry, functional genomics, high-throughput screening (HTS) and cell supply, has been substantially scaled up in recent years by ever increasing levels of automation. The development and deployment of automated systems in pharmaceutical and agrochemical discovery facilities show that the technology and automation is, in many ways, the easiest problem to solve. The real difficulties in scale-up lie elsewhere. The Drug Discovery Factory not only brings together the automated functional elements of the drug discovery process but also integrates them with the logistics, management and control into a single industrialised process.
The pressures to change the drug discovery process are the familiar ones of escalating costs, reducing profits and the need to bring drugs to market faster. For example, the huge growth in screening has been driven by this background. But the emphasis has been on automation with ever increasing potential throughputs. What is apparent from companies' results is that actual throughput is at least one order of magnitude less than the potential capability offered by current automated systems. The explanation for this comes later, but as an initial thought experiment, imagine how much quicker you would get to work if your car had a top speed of 1000 kilometres per hour? Clearly potential speed is not the most important issue.
AUTOMATING THE PHARMACEUTICAL INDUSTRY
The pharmaceutical industry tends not to look outside its own walls when it comes to process automation. In many ways it seems to believe that it has little to learn from other industries. However, when it comes to automation and related systems, much can be learnt from industries whose survival depends more closely on effective automation. For example, the car industry has always been at the forefront of automation technology.
Right back to Henry Ford the car industry has led the way in applying new production technology to provide substantial changes in product cost and quality. It is the largest single user of robots, but it is important to understand that robots are no longer used as a means of eliminating the cost of manual labour. The prime reasons for using robots today are the process quality and consistency they provide and, most importantly, the very high levels of process flexibility they can give as part of a computer integrated manufacturing system. Saving direct cost is no longer the critical factor. For more than twenty years in automotive manufacturing the rule has been ‘automate or die’. Automation is not only critical to competitive success but, done properly, it is also highly capital intensive. And the car companies have got good at it because they had to!
So where are the parallels between car manufacture and drug discovery? Car production is usually associated with long runs of a well-defined standard model with perhaps a major design change only every two years. This is, apparently, not much like the dynamic and uncertain world of drug discovery. But there are two myths relating to the Model T Ford production line which must be dispelled: any colour you like as long as it's black, and a thirty year product life. This is no longer the case because flexibility is designed into the product and process. In fact most cars today are built to order in about a week. Every car on the production line can be very different to the previous one and there are weekly changes to the product or process design. This is an extremely sophisticated system in which the so-called ‘Simultaneous Engineering’ of the product and process is absolutely essential. For drug discovery this means that a process is developed from the start to make best use of the factory's capability. Robust syntheses and assays, like robust designs, thrive in the hands-off environment of a modern production line.
Semiconductor production is another example worthy of the same comparison. Again, it is highly automated but for different reasons. The product and the process are essentially one and the same. Designing a chip with 10 million transistors is conceptually simple but it is totally useless if the precision of the silicon processing technology cannot make it. Again, product designs change frequently and production is massively parallel, i.e. there can be many hundreds of entirely different products being produced at the same time and using the same machines. In essence, this is what modern manufacturing technology is about: many different products at the same time on the same machines. All processes in semiconductor production are automated; not least to avoid the particulate contamination generated by people. The product is transferred between each process module by flexible systems to give a wide choice of process routes and sequences. Automation is inevitable here and the whole scheme has to be highly integrated with flexibility built in.
Again, the parallels with drug discovery include the need for extremely high levels of capital investment. Each process module is highly complex and supplied by a specialist vendor. People have little direct role in the process. Most critically, in contrast with cars, a chip's (or an assay's) production status cannot be established just by looking at it. Planning and tracking the work in progress depends totally on the computer system. For different reasons, cars and semiconductors have to be made this way. Sooner or later (and probably sooner) drug discovery will have to operate at such a scale that the same rules apply. Robust, integrated production lines, with built in flexibility and management to match will be inevitable.
Most people assume that labour reduction and other savings in direct costs justify automation. This is totally incorrect. Flexible automation is concerned with giving the customer whatever they want, whenever they want it and at a price they can afford, while controlling all the indirect costs of the process. If any version of a car can be built to order in a week, not only is the customer pleased with the choice and speed but also the manufacturer does not need to have the large inventories and costs of finished goods.
THE DRUG DISCOVERY FACTORY
Drug discovery must now be seen as an industrial scale process. At the process module level, several companies offer exciting new automated systems for compound storage, functional genomics, automated and combinatorial chemistry, ultra HTS and cell supply. Most major pharmaceutical companies have some activity in each of these areas but usually each one is in a separate department and is not yet part of a planned and fully integrated solution. Each process bottleneck is then addressed in a piecemeal way and the bottleneck is then moved to another department. Bottleneck chasing is a well-researched phenomenon in advanced production control.
The potential for automation is, however, not just at the process module level. It lies in the management and integration of all the stages in a multi-module complex process. This requires design, simulation and implementation techniques which are very familiar in other industries but rarely used in research. Running and staffing these complex facilities is not a peripheral issue. It is absolutely critical to achieving their true potential and realising the benefits of quality, cost, throughput, flexibility and choice.
It is very easy to regard drug discovery scale-up as being achieved by just doing more of the same. Unfortunately however it is well known in ‘catastrophe theory’, that there comes a point at which more throughput can only be achieved by a fundamental change in the process complexity (Figure 1). A further well known industrial phenomenon is that, even when all the necessary proven modules are in place and working properly, still only half of the potential benefit will be obtained unless all modules are totally integrated (Figure 2). Therefore, with respect to HTS, assay performance should be measured by the total time it takes to introduce a new assay, run it and generate results, not by the potential throughput of the separate process modules. In drug discovery, as elsewhere, most of the total process time is taken up waiting for something to happen — remember the Ferrari sitting at the traffic lights.
As process complexity is increases in line with increasing throughput, a point of chaos will arise at which no further increase in capacity can be achieved without a fundamental change in the process.
Even when all the necessary proven modules are in place and working properly, still only half of the potential benefit will be obtained unless all modules are totally integrated.
But here's the rub: clearly different kinds of people are needed to work in factories, different management methods are required and the general ethos is quite different from laboratories. To an extent these contrasts are obvious. The issue in drug discovery scale-up is that the numbers require industrial scale thinking but the current organisational cultures and staffing are strongly research oriented. It is clear that these are very uncomfortable companions. Change from a laboratory ethos which is exploratory, unconstrained, intellectually driven and uncertain in outcome to a production culture which is structured, rational, planned and with a predictable outcome is a major conceptual task. If drug discovery has to be industrial in scale, it should be a factory in practice. It must, therefore, have certain outline requirements on its performance. It must be responsive, incorporating new scientific methods and technologies, whilst being totally integrated. Processes must be designed for manufacture with documented procedures developed externally and machine friendly systems. In summary, the Drug Discovery Factory should be able to carry out ‘any assay on any compound, at any time, with time from request to result of less than a week and no mistakes’.
Let's take a walk through how the factory design might map on to that simple statement of requirements. A key principle is that the Drug Discovery Factory needs to be separate from the traditional culture and environment of a normal drug research campus. It may only be just across the road, but it can only succeed with a very different and distinct culture, systems, management and staffing. This also means that it can be commissioned and fully integrated quickly and without disruption to the existing discovery programmes. It can also be quicker and cheaper to build as there is no need for marble edifices and power architecture.
The five main process departments of automated dispensary and inventory management, functional genomics, automated/combinatorial chemistry, automated cell production and ultra high throughput screening could each run their own high capacity, high reliability automation systems but in response to the mutual needs of the overall factory. These are the visible elements and the process flows but the real magic lies elsewhere.
Superimposed on the robotics and equipment is the computing, software and data management needed not only to inject new assay methods and collect the data but also to manage, synchronise and integrate the entire process (Figure 3). The research perspective often assumes this type of software is just LIMS. In many ways that is the easy part that handles data capture. Without the additional synchronising systems of inventory management and tracking, MRP, scheduling, process control, maintenance, data analysis and processing, the process cannot and will not work.
Without the additional synchronising systems of inventory management and tracking, MRP, scheduling, process control, maintenance, data analysis and processing, the main process departments of the Drug Discovery Factory cannot and will not work.
The earlier diagram showed the simple department blocks and workflow. In reality, apparently simple processes are always more complex when you take into account that it is a process which keeps changing and the factory design must allow for this complexity and uncertainty. Complexity and flexibility can co-exist, but only if this requirement is specified from the outset.
No one would suggest that the medicinal chemist who developed a drug be best qualified to produce the tablets. The same logic applies to the industrialisation of drug discovery. The process cannot be dependent on the art of a ‘green fingered’ technician. It has to be robust, consistent and repeatable. It also has to operate round the clock. This requires machines that can run without breakdown for weeks not minutes. Most importantly, the science has to be done away from the factory.
SUMMARY
The Drug Discovery Factory is about discovering new drugs faster without placing any unreasonable restraints on the science or assay platforms, it is not about removing people or eliminating the intellectual elements of the discovery process. The prime requirement is that, as a user, you get whatever you want, whenever you want it together with the following benefits:
It is a distinct entity (Figure 4) that can be supplied and operated as a turnkey package. The customer does not need to understand the internal interfaces and processes.
The Drug Discovery Factory is a distinct entity that can be supplied and operated as a turnkey package and staffed by factory workers. It can be sited in an industrial area away from the laboratory-based research and development functions.
The costs per test will be reduced, principally by reducing the indirect costs.
The intrinsic accuracy of the process and quality of the data will be improved because of consistency and reliability.
It is an almost inevitable consequence of the current projections of HTS growth.
It is not anti-science but it allows the intellectual process to operate separately from the HTS process with potential benefits to the quality of both.
To be clear, this is neither a dream nor, it is hoped, a nightmare. Much of this is well understood in very comparable situations in other industries. The successful realisation of the Drug Discovery Factory lies not only in technology and management but also in a cultural change to the way the pharmaceutical industry thinks.