Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

PSTT Vol. 1, No.

4 July 1998 research focus reviews

Artificial intelligence in
pharmaceutical product formulation:
knowledge-based and expert systems
Raymond C. Rowe and Ronald J. Roberts

The pharmaceutical formulation process is a highly specialized task • ‘An expert system is a computer program that
requiring specific domain knowledge and often years of experience. draws upon the knowledge of human experts
captured in a knowledge base to solve prob-
Knowledge-based systems, expert systems, neural computing and lems that normally require human expertise’2
machine-learning derived from research into artificial intelligence
Because of their emphasis on knowledge, the
can assist the efficient formulation of products and therefore increase
terms ‘expert system’ and ‘knowledge-based sys-
productivity, consistency and quality. The authors provide an introduc- tems’ are often used interchangeably. If any dif-
tion to the technology of knowledge-based and expert systems. ference does exist, it is in how their input knowl-
edge is acquired. In expert systems input
knowledge is acquired, in the main, from human
▼ In all cases of product formulation – whether experts, whereas in knowledge-based systems it
Raymond C. Rowe* and
Ronald J. Roberts
it be for tablets, capsules, parenterals or any one is usual for input knowledge to be acquired
Zeneca Pharmaceuticals of the numerous other dosage forms – in generic through non-human interaction, such as through
Alderley Park terms the process is the same, beginning with information systems and databases.
Macclesfield the presentation of some form of product speci- In their simplest form, both systems have three
UK SK10 2NA fication and ending with the generation of one major components: an interface, monitor and
*tel: ⫹44 1625 513112
or more formulations that will fulfil the require- keyboard allowing two-way communication be-
fax: ⫹44 1625 512381
e-mail: ments. In many cases the specification has poten- tween the user and the system; a knowledge base
Ray.Rowe@alderley.zeneca.com tially conflicting requirements and thus it is left in which all the knowledge pertaining to the do-
to the highly skilled and experienced formulator main is stored; and an inference engine in which
to balance these and produce the optimum com- the knowledge is extracted and manipulated to
promise formulation. solve the relevant problem. Inferencing strategies
It is not surprising, therefore, that product may be either forward chaining, which involves
formulation has become a popular target for the the system reasoning from data and information
development of expert systems due to the poten- obtained by consultation with the user to form a
tial pay off of a system that can create formu- hypothesis, or backward chaining, which in-
lations with little or no human intervention. volves the system starting with a hypothesis and
then attempting to find data and information to
Technology prove or disprove the hypothesis. Both strategies
There is a wide range of views of what defines are included in most expert systems.
an expert system. Examples of widely used defi- Knowledge in any domain takes the form of
nitions include: facts and heuristics; the former being valid, true
and justifiable by rigorous argument, the latter
• ‘An expert system is a knowledge-based sys- (often referred to as ‘rules of thumb’) being the
tem that emulates expert thought to solve sig- expert’s best judgement in any particular circum-
nificant problems in a particular domain of stance and hence justifiable only by example.
expertise’1 Associated with these are the terms ‘data’ and

Copyright ©1998 Elsevier Science Ltd. All rights reserved. 1461-5347/98/$19.00. PII: S1461-5347(98)00042-X 153
reviews research focus PSTT Vol. 1, No. 4 July 1998

Box 1. Levels and categories of knowledge A technique that is often used in the acquisition process is
the rapid prototyping approach. In this approach, the knowl-
• Shallow – Surface-level, specific, heuristic edge engineer builds a small prototype system as early as
• Deep – Complete, integrated, conceptual possible and this is shown to both the expert and user, who can
• Explicit – Conscious, able to be articulated then suggest modification and additions.This system grows in-
• Tacit – Implicit in expert’s ability to solve problems crementally as more information and knowledge is gained, and
• Declarative – Descriptive representation of facts this methodology has been used successfully in the develop-
• Procedural – Detailed set of rules ment of systems for pharmaceutical formulation.
• Meta – Knowledge about knowledge Once acquired, there are many ways of representing the
knowledge in the knowledge base (Box 2). It is probable that
the most common methodology is the production rule; this ex-
‘information’, the former referring to facts and figures, the lat- presses the relationship between several pieces of information
ter being data transferred by processing such that it is mean- by way of conditioned statements which specify sections under
ingful to the recipient. Knowledge can therefore be regarded as certain sets of conditions.
information combined with heuristics and rules. There are
many levels and categories of knowledge (Box 1) and it is the • IF (Condition 1)
objective of the knowledge engineer to acquire or elicit this • AND (Condition 2)
knowledge and organize it into a computer-readable format. • OR (Condition 3)
Knowledge acquisition is probably one of the most difficult • THEN (Action)
stages in the development of an expert system. It is both time • UNLESS (Exception)
consuming and tedious as well as being expensive and often • BECAUSE (Reason)
difficult to manage. However, it is a necessary element in the
building of an expert system and, if done well, will undoubt- Each rule implements an autonomous piece of knowledge
edly lead to systems of potential use.The basic model of knowl- and is easy to understand. Unfortunately, complex knowledge
edge acquisition is one of a team process whereby the knowl- can require large numbers of rules, causing the system to be-
edge engineer mediates between the expert(s), the users and come difficult to manage. The decision over which method of
the knowledge bases.The knowledge engineer must acquire or knowledge representation should be adopted is dependent pri-
elicit knowledge from not only the expert(s) but also all the marily on the complexity of the domain. In a survey of 50 ex-
other potential sources; these include written documents (re- pert systems in the UK in 1987/1988, it was found that, al-
search reports, reference manuals, operating procedures policy though over half the systems were developed using production
statements) and consultants, users and managers. In the case of rules, nearly a sixth used multiple approaches3.
experts, in the main knowledge is acquired through face-to- Expert systems can be developed using either conventional
face interviews. This process is tedious and can place great de- computer languages, special purpose languages or with the as-
mands on both the expert and knowledge engineer but re- sistance of development shells or toolkits. Conventional lan-
quires little equipment (e.g. tape recorder or note book), is guages such as PASCAL and C have the advantages of wide
highly flexible and often yields a considerable amount of use- applicability and full flexibility to create the control and infer-
ful information. At all times, there must be an empathy be- encing strategies required. They are also well supported and
tween the participants and, in many cases, it is helpful to have easy to customize. However, considerable amounts of time and
two knowledge engineers present at the interview. effort are needed to create the basic facilities.
Specialized languages such as LISP (a recursive language and
the primary one for artificial intelligence research), PROLOG (a
Box 2. Methods of knowledge representation language based on first order predicate logic) and SMALLTALK
(an object-orientated language) have been used extensively in
• Production rules the development of expert systems because they share the ad-
• Frames vantages of applicability and flexibility of the conventional
• Semantic networks languages but are also faster to implement.
• Decision tables Expert system shells and toolkits are sets of computer pro-
• Decision trees grams written in either conventional or specialized languages
• Objects that can form an expert system when loaded with the relevant
knowledge. They compromise on applicability and flexibility

154
PSTT Vol. 1, No. 4 July 1998 research focus reviews

27 April 1989 in the London Financial Times7; this was followed


Table 1. Published applications of pharmaceutical product
closely by an article in the autumn of the same year by Walko8.
formulation expert system
Both refer to the work then being undertaken by personnel at
Company/Institution Domain Development tool ICI (now Zeneca) Pharmaceuticals UK and Logica UK, to de-
velop an expert system for formulating pharmaceuticals using
Cadila Laboratories (India) Tablets PROLOG PFES. Since that time, several companies and academic institu-
University of Heidelberg Aerosols C/SMALLTALK tions have reported their experiences in this area (Table 1). It
Tablets should be noted that those systems listed have been described
Capsules
in the open literature; other companies/institutions have de-
IV injections
veloped systems for similar domains but many have been
University of London/Capsugal Capsules C
Sanofi Research Capsules PFES
reticent to report their experiences.
Zeneca Pharmaceuticals Tablets PFES
Parenterals Cadila System
Film coatings This system was developed by personnel at Cadila Laboratories
Boots Company Topicals PFES (Ahmedabad, India) and has been designed to formulate
tablets for active drugs based on their physical (e.g. solubility,
but allow more rapid development. Many offer basic facilities hydroscopicity), chemical (e.g. functional groups) and bio-
such as the means to prepare and store knowledge as a set of logically inter-related (e.g. dissolution rate) properties9. The
rules and to make deductions by chaining the rules together in system first identifies the desirable properties for optimum
an inferential process. compatibility with the drug, selects those excipients that have
Shells differ in their secondary characteristics, such as user the required properties and then recommends proportions
interfaces, operating speeds, the method of knowledge represen- based on the assumption that all tablet formulations comprise
tation and the associated algorithmic and arithmetic computa- at least one binder, one disintegrant and one lubricant. Other
tional facilities. One such shell of specific importance in product excipients such as fillers or glidants are then added as required.
formulation is Logica’s Product Formulation Expert System Knowledge acquired through active collaboration with ex-
(PFES, Logica UK Ltd), a reusable software kernel and associated pert formulations is structured as decision tables with derived
methodology developed from underlying research and con- production rules. The system written in PROLOG is menu-
ducted by a consortium of Shell Research, Schering Agrochemi- driven and interactive with the user who is prompted to enter
cals, and Logica UK, under the UK Alvey Programme sponsored all the known properties of the new active drug. When proper-
by the UK Department of Trade and Industry between 1985 and ties have descriptors (e.g. functional groups, solubility, etc.)
1987. PFES was designed specifically for building formulation these are displayed for selection by the user. The system then
expert systems. Its capability is generic, providing knowledge suggests a formulation with alternatives and explanations (if
representations common to most product formulation tasks and required). An example formulation for a dose of 500 mg of
thus allowing new applications to be developed rapidly and ef- paracetamol is shown in Table 2. The filler is unnamed, but it
fectively. It provides a decision support framework consisting of can be assumed that it will not be lactose because the relevant
two levels: a task level that contains the distant problem-solving production rule indicates that there would be an interaction
steps involved in creating a formulation, and a physical level between it and the secondary amine group on the drug.
that contains specific knowledge about the properties of the The prototype system, when first implemented, had 150 rules
ingredients and processes involved4,5. but this has rapidly expanded to over 300 rules to increase reli-
The choice of development method used depends, inter alia, ability. It took 6–7 months to develop and is reported to have re-
on the domain, the type of knowledge to be encoded, the hard- duced the development time for a new tablet formulation by 35%.
ware platform and the end-user environment. In 1991, a survey
of expert systems showed a distribution of methods of 11% Table 2. An example of a tablet formulation for paracetamol
conventional languages, 23% specialized languages and 66% as generated by the Cadila System
shells and toolkits6. All the methods have been used for phar-
maceutical product formulation (Table 1). Paracetamol 500.0 mg
Unnamed filler 20.0 mg
Applications Pre-gelatinized starch 43.7 mg
Sodium starch glycolate 5.0 mg
The first recorded reference to the use of expert systems in
Stearic acid 2.5 mg
pharmaceutical product formulation was by Bradshaw on

155
reviews research focus PSTT Vol. 1, No. 4 July 1998

Galenical Development System, Heidelberg a statistical design to optimize the formulation quantitatively,
This system, developed by personnel in the Departments of the specification of the excipients used, recommended tests to
Pharmaceutics and Biopharmaceuticals and of Medical validate the formulation and a complete documentation of the
Informatics at the University of Heidelberg (Germany), has decision process.
been designed to provide assistance in the development of a The system is unique in that its knowledge base is broad; it
range of formulations (e.g. aerosols, tablets, capsules, intra- contains information on a large number of excipients; a data-
venous injections), starting from the chemical and physical base of marketed formulations from Germany, Italy, Belgium,
properties of an active drug. The project was initiated in France and the USA; a database on literature references associ-
199010 and extensively revised and enhanced in the in- ated with the formulation of hard gelatin capsules; experience
terim11,12. Originally implemented in C, the system has re- and non-proprietary knowledge from a group of international
cently been upgraded using SMALLTALK. Knowledge represen- experts; and the results from statistically designed experiments
tation is by means of objects, frames and production rules. on capsule formulation. The data is frequently updated and the
Because of the complexity of the overall process, the devel- system has a semi-automatic learning tool that monitors users’
opers have adopted an approach whereby the system focuses on habits regarding the use of excipients. These, together with the
one problem and its associated subset of specification and con- results from user questionnaires, are analysed by the expert
straints at a time. These distinct development steps are worked system founder group providing the background to further
through in succession, care being taken that any solution enhancements.
should not violate any constraints from previous steps.
The output of the system is a formulation together with pro- Sanofi System
duction method, recommended packaging and, where appro- This system, developed by personnel at the Sanofi Research
priate, predicted product properties. Explanation of the strategy Division, has been designed to formulate hard gelatin capsules
used, formulae and rules fired as well as text and literature ref- based on specific preformulation data on the active drug14.
erences can be produced on request. An example formulation The system, implemented using PFES, generates one first-pass
for a 150 mg dose of griseofulvin in a hard gelatin capsule is capsule formulation with as many subsequent formulations as
shown in Table 312.The system was introduced for field trials in desired to accommodate an experimental design. Knowledge
December 1996. acquisition was by interview and structured using objects,
frames and production rules.
University of London/Capsugel System The user is first prompted to enter specified preformulation
This system, developed by personnel at the School of Pharmacy, data on the active drug and the system recommends a formu-
University of London (UK), supported by Capsugel together lation with relevant capsule size and production information.
with personnel from the Universities of Kyoto (Japan) and An explanation log, providing details of the decisions and rea-
Maryland (USA) has been designed to aid the formulation of soning used by the system, are also recorded. An example of a
hard gelatin capsules13.The system, implemented in C, uses de- formulation of 100 mg of naproxen is shown in Table 4.
cision trees and production rules for knowledge representation.
The system is fully interactive with the user and has a ques- Table 4. An example of a capsule formulation for naproxen
tionnaire to assist in collecting all necessary input data. From as generated by the Sanofi System
these data the system uses a variety of methods to predict prop-
Naproxen 100 mg
erties of mixtures of the new drug and various excipients and
Lactose (hydrous) 224 mg
recommends a formulation with any necessary powder pro-
Microcrystalline cellulose (PH105) 060 mg
cessing and filling conditions. In addition, the system provides
Sodium lauryl sulphate 004 mg
Talc 012 mg
Table 3. An example of a capsule formulation for
griseofulvin as generated by the Galenic Development Capsule size No. 0
System, Heidelberg, Germany

Griseofulvin 150.0 mg Zeneca systems


Microcrystalline cellulose (PH102) 199.2 mg Since April 1988, three systems have been developed by person-
Magnesium stearate 3.5 mg nel at Zeneca Pharmaceuticals15–17. These systems, designed for
formulating tablets, parenterals and film coatings, respectively,
Capsule size No. 1
have all been implemented using PFES. Each was developed using

156
PSTT Vol. 1, No. 4 July 1998 research focus reviews

relatively high investments of time and resources are often


Table 5. Example of tablet formulation for a model drug as
generated by the Zeneca System18 needed for their development, it is imperative to identify and
objectively assess their benefits. In a large survey in 1989 in-
Drug A 50.0 mg 150.0 mg volving nearly 450 responses from a wide variety of organiza-
Lactose monohydrate 166.9 mg – tions and types of users, several key benefits were identified20.
Dicalcium phosphate dihydrate – 165.7 mg Overall, the most important benefits were those concerned
Croscarmellose sodium 4.8 mg 7.0 mg with the accuracy of the decision making, improved problem
Polyvinyl pyrrolidone 4.8 mg –
solving and quality/accuracy of work (Fig. 1).The least impor-
Hydroxypropyl methylcellulose – 7.0 mg
tant benefits were those concerned with the reduction in staff
Sodium lauryl sulphate 0.7 mg 1.1 mg
Magnesium stearate 2.4 mg 3.5 mg
numbers or using less skilled staff.
Thomas21 has attempted to capture the overall business im-
Tablet diameter 8.0 mm 10.0 mm pact of expert studies in terms of quality availability and consist-
ency of expertise available. Expert systems should enable man-
agers to begin to manage expertise as a corporate asset leading
two experts, one with extensive heuristic knowledge, the other to an increased competitive edge, improved risk management,
with extensive research knowledge, and structured using objects, increased revenue, decreased costs and increased profitability.
frames and production rules by specialized consultancy support. Such is the overall picture of the benefits of developing an
Each system is fully interactive with the user, requiring expert system. It is more pertinent to discuss those found by
specified input data on the active drug. The systems then rec- the users of systems in pharmaceutical formulation.
ommend formulations with predicted properties (e.g. com-
paction properties for tablets, tonicity and storage life for par- Knowledge availability and protection
enterals, opacity and cracking potential for film coatings). Each The developers of the University of London/Capsugel System
system contains ‘Help’ routines providing both on-line help in have reported the benefit of being able to use knowledge from a
the use of the system and the rationale behind the adoption of broad base13. Others9,14 have reported the benefit of the prompt
the specified rules/formulae used. The user is able to browse availability of information and the rapid access to physical and
the knowledge trees at will but is unable to edit them without chemical data of both drugs and excipients reducing the time
privileged access. Explanations for any recommendations made spent searching literature. Rowe15 has reported the existence of a
by the systems can be accessed easily if required. The tablet coherent durable knowledge base not affected by staff turnover.
formulation system is the most frequently used expert system
and is now an integrated part of the development strategy in Consistency
this domain. An example of a tablet formulation for a model All systems generated robust formulations with increased cer-
drug recommended by the system is shown in Table 5. tainty and consistency.This was seen as a distinct benefit where
regulatory issues are important.
Boots System
Although not strictly developed for pharmaceutical formu-
lation, this system has been included since it is the only one Accuracy of decision making
known for formulating topicals. Developed in 1990 by person- Improved problem solving

nel at the Boots Company, the system won second prize in Accuracy of work
Quality of work
the UK Department of Trade and Industry Manufacturing
Cost effectiveness
Intelligence Awards in 199119. The system was developed in
Increased output
the first instance to assist in the formulation of sun oils but has
Reduced skill level
been rapidly extended to cover creams and lotions. Knowledge
Reduced skilled staff
was acquired from senior formulators using interviews and
Increased throughput
structured as objects, frames and production rules.
Reduced staffing
Implemented in PFES, this is the only system in which the de-
velopers have given details of costings and quantitative benefits. 0 5 10 15 20 25
Total mentions (%)
Benefits Figure 1. A survey of key benefits for expert systems as reported by
Although there is a great deal of interest in expert systems, Dewar20.
there is still much uncertainty regarding their benefits. As

157
reviews research focus PSTT Vol. 1, No. 4 July 1998

Training No budget
All systems have been used to provide training for both novices No management support

and experienced formulators. In topical formulation, use of the No benefits


No management awareness
expert system has enabled experienced formulators to experi-
No expert
ence unfamiliar new excipient combinations19. No knowledge engineer
No financial gain
Speed of development No confidence
Reduction in the duration of the formulation process has been Prefer conventional
widely reported9,15,19. It has been reported that formulators No technical support

working in the topical domain can now produce a formulation No user

in 20 minutes that might otherwise have taken two days to 0 5 10 15 20


achieve19. A 35% reduction in the total time needed to develop Mentions (%)
a new tablet formulation has also been reported9.
Figure 2. A survey of issues that prevented the implementation of
expert systems as reported by Dewar20.
Cost savings
Cost savings can be achieved by not only reducing the develop-
ment time but also by the more effective use of materials.
Several users have reported a benefit in being able to plan the advantages of expert systems (Box 3), and experience in the
purchase and stocking of excipients and to reduce the size of product formulation domain would suggest that these be
raw material inventories9,19. adopted as a screening process for assessing the technical
feasibility of an application.
Freeing experts Because the development of an expert system is likely to be
The implementation of expert systems in product formulation high profile and failure will be very visible, a poor choice of
has inevitably allowed expert formulators to devote more time application can have disastrous consequences, not least in the re-
to innovation14,19. In topical formulation it has been reported jection of the technology. Attempting to develop an application
that the use of an expert system has released approximately that will solve all problems in the first instance can be just as bad.
30 days of formulating time per year per formulation19. It is better to limit the scope of the application at the outset or
sub-divide the domain into small manageable modules and de-
Improved communication velop each in turn; this has been done in the Galenical Develop-
Several users have reported that expert systems promote com- ment System (Heidelberg, Germany)10–12. Experience has shown
munication and discussion amongst not only experts but also that, in general, the quality of the performance of the developed
managers, enabling them to identify critical areas requiring re- system is more important than its scope. However, trivial appli-
search and/or rationalization9,14,19. cations should not be developed because these can rapidly lose
credibility and support from both experts and management.
Of all the systems developed for pharmaceutical formulation
only one of those provided costings and had undertaken a
cost–benefit analysis. Wood19 has reported that the expert sys-
tem developed by the Boots Company cost £25,400 consisting
Box 3. Features of applications where expert
of £10,400 for hardware and software, £6000 for consultancy
systems may provide advantages
and £9000 for experts’ time.With an estimated annual cost sav-
• Primarily symbolic in nature
ing in the region of £200,000 this system delivered a payback
• Rule base and heuristics available
of approximately three months. • Neither too easy nor too difficult
• Manageable size
Issues • Availability of at least one cooperative and articulate expert
There are several issues that have reduced levels of general im- • Non-optimal results tolerable
plementation of expert systems (Fig. 2). Without the potential • Of practical importance
to deliver real business benefit, management cannot be • Part of essential activity
expected to support resource projects and hence the identifi- • Occur in a stable environment
cation of the most suitable application is an essential pre- • Meet real business need
requisite. There are many features that indicate possible

158
PSTT Vol. 1, No. 4 July 1998 research focus reviews

Managers have a pivotal role in the development of expert has, in the main, changed to a mood of enthusiastic partici-
systems. They do not need to have an in-depth understanding pation. It is unlikely that expert systems will ever replace for-
of the technology but must set the framework within which mulators but as a decision-support tool they are invaluable and
the developers work.The management role is crucial to the suc- deliver many benefits.
cess of the project and, as a result of extensive studies, it is now
known that organizations that are successful in the develop- References
ment of expert systems are those whose management play an 01 Sell, P.S. (1985) Expert Systems – A Practical Introduction, Camelot Press
active role in coordinating and supporting their initiatives. Ex- 02 Partridge, D. and Hussain, K.M. (1994) Knowledge-Based Information Systems,
pert systems projects are no different from any other project McGraw Hill
and can be planned and managed using common prac- 03 O’Neill, M. and Morris, A. (1989) Expert Systems 6, 90–91
tices6,22,23.This does not mean that they are well managed. In a 04 Turner, J. (1991) Manufacturing Intelligence (issue 8), 12–14
survey of expert systems applications from all UK business sec- 05 Bentley, P. (1998) in Intelligent Software for Product Formulation (Rowe, R.C. and
tors, it was found that only 20% of respondents had used any Roberts R.J.), pp 27–41,Taylor and Francis
project planning/management systems20. 06 Edwards, J.S. (1991) Building Knowledge-Based Systems: Towards a Methodology, Pitman
An issue that must be managed sensitively at the outset is 07 Bradshaw, D. (1989) Financial Times, 27 April, 26
that of the role of the expert and users once the system has 08 Walko, J.Z. (1989) Innovation (issue 18), 24
been implemented. Although in the surveys (Fig. 1) reduced 09 Ramani, K.V., Patel, M.R. and Patel, S.K. (1992) Interfaces 22, 101–108
staff levels have been seen as one of the least important benefits 10 Stricker, H. et al. (1991) Pharm. Ind. 53, 571–578
of implementation, it can still be an issue with some individ- 11 Stricker, H. et al. (1994) Pharm. Ind. 56, 641–647
uals. It is inevitable that there will be a change in the job de- 12 Frank, J., Rupprecht, B. and Schmelmer,V. (1997) IEEE Expert 12, 40–48
scription of both the expert and users and this may or may not 13 Lai, S. et al. (1996) Pharm.Technol. Eur. 8(9), 60–68
result in a change in the opportunities for career development. 14 Bateman, S.D. et al. (1996) Pharm.Technol. 20(3), 174–184
This can result in a lack of cooperation by both users and ex- 15 Rowe, R.C. (1993) Manufacturing Intelligence (issue 14), 13–15
perts in both the development and implementation. 16 Rowe, R.C. et al. (1995) PDA J. Pharm. Sci.Technology 49, 257–261
17 Rowe, R.C., Hall, J. and Roberts, R.J. Pharm.Technol. Eur. (in press)
Conclusion 18 Rowe, R.C. and Roberts, R.J. (1998) Intelligent Software for Product Formulation,
Although there are many issues surrounding the development Taylor and Francis
of expert system applications, the majority of these can be 19 Wood, M. (1991) Lab. Equipment Digest December, 17–19
minimized by good management. The technology is mature 20 Dewar, J. (1989) Systems International 17(7), 12–14
and has proved its ability to generate real business benefit. It 21 Thomas, M. (1991) Manufacturing Intelligence (issue 7), 6–8
should now be considered routinely as one of the methods for 22 Guida, G. and Tasso, C. (1994) Design and Development of Knowledge-Based Systems
increasing competitiveness.Within the pharmaceutical product from Life Cycle to Methodology, John Wiley & Sons
formulation domain, early scepticism amongst potential users 23 Shaw, R. (1990) Expert Systems Opportunities – Case Studies, HMSO

In the July issue of Drug Discovery Today


Editorial – Trends in drug discovery technologies Future considerations in HTS: the acute effect of
Bob Wallace chronic dilemmas
Update – latest news and views Matthew A. Sills

High-throughput approaches to voltage-gated ion Miniaturization technologies in high-throughput


channel drug discovery screening: how fast, how small, how soon?
Jane Denyer, Jennings Worley, Brian Cox, Gary Allenby Jonathan Burbaum
and Martyn Banks Changes in scale in automated pharmaceutical
Tyrosine kinase assays adapted to homogeneous research
time-resolved fluorescence Willie Harrison
Alfred J. Kolb, Paul V. Kaplita, David J. Hayes, Young-Whan Monitor – new bioactive molecules, discovery technologies,
Park, Christine Pernell, John S. Major and Gérard Mathis emerging therapeutic targets

159

You might also like