Fast Formulations

By Stephen Warde, Molecular Simulations Inc.

Advanced design software and information management tools are accelerating formulations development. This technology represents a practical application of knowledge management.

Formulation sounds easy. Take well-known ingredients and mix them to make something with better properties. Make tablets that are air- and temperature-stable. Make a cough mixture that costs less than existing formulas, but still tastes great. Deliver a new drug at the same dose every time. All very straightforward. However, did your tablet just get too large for patient comfort? Can you make varieties of your cough mixture to match tastes in different countries? Will your new drug cost too much to transport and deliver? For any of these problems, where do you begin if you have hundreds of potential ingredients and millions of ways to combine them?

Formulation: Art Under Pressure
Trends in Industry and IT
MSI's Formulations Consortium
Supporting the Formulations Process
Why a Consortium?
The Knowledge Perspective

Formulation: Art Under Pressure (Back to Top)
Formulation is critical to adding value in chemical, pharmaceutical, and consumer product businesses. It is the science by which these businesses develop new and improved products using existing technologies and materials. But not only must they produce better properties, they must also understand regulatory constraints, customer requirements, cost implications, and manufacturing issues. These often conflict and must be balanced against each other. A typical formulation problem has dozens of such variables. Good formulators work around some of this complexity through their experience. Unsurprisingly, formulation is often viewed as an art, the domain of the gifted expert carrying an invaluable knowledge base in his or her head. Experiment is often conducted on a trial-and-error basis.

If formulation is difficult given perfect information about its inputs and outputs, how much harder is it in the real world? In a typical pharmaceutical company, the formulator works in product development. Data on product usage conditions and customer satisfaction reside in the marketing department. The latest information on regulations affecting the product comes from half a dozen different business units. Purchasing departments provide materials costs that change daily. Even testing is often done in separate production or analysis groups. Rarely is all of the critical information at the formulator's fingertips. Where corporate databases provide it, they are often designed for purposes other than product development.

Trends in Industry and IT (Back to Top)
A number trends in the pharmaceuticals industries magnify these problems. Leading companies now run global operations in global markets, so formulators must respond to diverse customer-bases and collaborate with colleagues worldwide. Intense competitive pressures mean that more products must be generated faster, and at lower cost. The ability to rapidly meet very specific customer needs is increasingly important, making trial-and-error an even less secure basis for competitive advantage. This business environment is driving a spate of mergers, posing new formulation problems and introducing issues of data integration and compatibility. Finally, companies face an increasing degree of retiring or departing expertise. To protect their competitive position they must find a way to retain and share the knowledge in expert formulators' heads.

Computers should help to meet these challenges. Computational methods are a well-established means for rationalizing product design. Information technology (IT) now enables instant access to data, information, or knowledge, wherever it originates. Yet such technology has yet to dramatically impact formulation design. Systematic efforts to look for better products, for example by combining experimental data with statistical models, are under-utilized. Much of the software that is available only makes matters worse, principally because it has been designed for other types of users. Statistical packages, for example, employ jargon such as "descriptors" and "dependent variables", rather than using language familiar to formulators. Nevertheless, if these barriers can be overcome, the IT infrastructure required to support an effective formulation design system exists in most research organizations.

MSI's Formulations Consortium (Back to Top)
The development and deployment of such system is the aim of MSI's new Formulations Consortium. This group of commercial research organizations will create an integrated software toolkit to support "best practice formulation". This term encapsulates processes and practices designed to minimize time-to-market for new or remodelled formulations, to guarantee superior price/performance and quality, and to capture and share knowledge. The toolkit will maintain and mine databases containing relevant information from marketing, research, development, testing, and production. It will provide methods to design and optimize formulations and it will support design of experiments and analysis of their results. The first member of the Consortium is Unilever. Unilever joined following one year of a contract development project with MSI. "The resulting software system helps our formulators to access critical data, share expertise, and accelerate key steps in developing detergents," explains Antoine Schlijper, Project Manager of Unilever's Laundry Research group.

MSI's experience in serving the pharmaceuticals and chemicals industries and in running four other industrial consortia will be critical to the success of the Formulations Consortium. This expertise will help to overcome the major obstacles to the systematic application of IT in formulation design. First, there is the need for a more rational approach. MSI has a 15-year record of providing software to rationalize scientific problems while supporting researchers' creativity. To achieve this, the company draws on a unique pool of chemical modeling technology. Second, there is a requirement for productive software. MSI has a proven ability to design tools specifically for scientists. The Consortium software will operate through a user interface designed in close collaboration with practicing formulators, using familiar terminology and methodologies. It will be fully compliant with Windows NT and Oracle databases.

Supporting the Formulations Process (Back to Top)
This software is known as FAST (Formulation Assisting Software Toolkit). FAST will aid formulation project teams within a framework offering support for the complete design process. This process usually begins with the marketing department specifying a project based upon customer or business requirements. Formulators then begin with a base formulation from a supplier, from historical data, or from an existing product. They modify this in an attempt to meet the specification, testing tentative formulations in the laboratory and pilot plant. Results are iteratively used to improve the formulation by seeking the best trade off among conflicting target criteria.

With FAST accessible from their PC, the formulator will begin by data mining in a formulations database. This stores information from previous projects, including ingredient properties, recipes, test results, costs, and usage conditions. Formulators will retrieve and re-use recipes with similar performance to the specification. They will also be able to retrieve correlation models that connect formulation properties with inputs to the recipe. Data analysis tools allow model-building from the database, enabling the formulator to maximize use of existing data. Most models will be based on macro-scale properties, such as ingredient composition. However, micro-scale variables that express the system's chemistry will enrich the description of some formulations. MSI has particular strengths in QSAR models that represent these variables. Tentative formulations can now be constructed on the computer and their properties estimated. Optimization methods may then be used to balance the recipe with respect to multiple objectives. The next step applies Design-of-Experiments (DOE) methods to select better starting points for testing and to help formulators to extract maximum information from the minimum number of experiments.

Results from these experiments feed back into the project database and may be used to refine the correlation models used in earlier steps—the process becomes a closed loop. The formulator continues around the loop until satisfied with the product. This systematic approach should require fewer iterations than trial-and-error. For every such project, the database stores all of the data collected and most of the information and knowledge gained. Because the system is installed enterprise-wide and integrated with standard PC tools, it is simple to share data with colleagues through the database and other media, such as e-mail, web pages, spreadsheets, presentations, and text documents.

Why a Consortium? (Back to Top)
Creating such a system needs existing technology to be creatively applied and extended. Its applications will cut across a wide range of products and industries. It will provide significant competitive advantages for early-adopters. The formulations system is configured to match each consortium member's business procedures, yet there is no sharing of proprietary information between members. Finally, it calls for a synthesis of software, systems, and scientific skills not readily available in many companies. These are the ideal ingredients for a pre-competitive consortium. MSI's consortium approach, as it has in other areas, will help to determine the best way to combine and implement technologies to maximize their impact on commercially significant problems. Member organisations typically join for three-year periods. Membership entitles researchers to use the consortium software, to provide input to software development, to attend regular meetings at which ideas are shared with peers, to receive dedicated application support, and to gain early access to new technology. Each consortium is supported by a panel of advisors who provide world-leading expertise in relevant technologies, as well as by MSI's own expert staff.

MSI's four current consortia average 25 member organizations. The newest, the Pharmaceutical Development Consortium has demonstrated MSI's ability to pioneer the use of computation in downstream areas of pharmaceutical research. MSI's Catalysis and Polymer Consortia are now both in their ninth year, indicating the long-term viability of the consortium concept. The Combinatorial Chemistry Consortium has successfully integrated simulation with informatics and data mining technology.

The Knowledge Perspective (Back to Top)
The Formulations Consortium continues the move from tools that create knowledge towards solutions that also manage that knowledge. IT alone is not knowledge management, but IT can create necessary conditions for knowledge management. These enabling conditions are what the Consortium offers. Knowledge is generated through innovative design tools. Concepts may be evolved using data mining and analysis. Knowledge is shared and re-used through the database. Antoine Schlijper at Unilever remarks, "The Consortium will create the software infrastructure required to support knowledge management in formulations design. As a leader in the practical application of knowledge management techniques, Unilever is delighted to be a founder member of this project."

Without such a system, the knowledge that has accumulated in a company may never be discovered, passed on, or used more than once. Formal knowledge management prevents this waste of resources. In their book If Only We Knew What We Know (Simon & Schuster, 1998), Carla O'Dell and Jackson Grayson report Chevron's calculation that it has saved over $650 million since 1991 by sharing best practices among managers in charge of energy use. A knowledge management program has helped Texas Instruments save more than $1 billion in its semiconductor plants. The formulation design process fits exactly into this mold. There is much expertise and information within companies, yet so little is audited, accessible, or used. There is rarely a formal mechanism for sharing best practice formulation design. The Formulation Consortium will help to change this situation.

In practice this means better drugs, paints, foods, detergents, chemicals, and plastics, matched to customer needs in less time, at lower cost, and meeting the relevant regulations.

For more information: Stephen Warde, Molecular Simulations Inc., 230/250 The Quorum, Barnwell Rd., Cambridge CB5 8RE, UK. Tel: +44 1223 413300. Fax: +44 1223 413301.