TECHNOLOGIES
DRUG DISCOVERY TODAY
Evolution in thinking and processes?
Andreas Sewing
Exploratory Medicinal Science, Pfizer Global Research and Development, IPC 654, Ramsgate Road, Sandwich CT13 9NJ, United Kingdom
Pharmaceutical R&D transforms scientific ideas into drugs on the market. Owing to the complexity and low overall success rate, Drug Discovery needs to be as much about science as about operational excellence. In vitro screening groups, underwriting early discovery from exploratory to candidate selection, are trying to combine the search for new scientific concepts with a production-like focus on logistics, reproducibility and delivery on time. Moving beyond high-throughput tech-nologies, we begin to ask how to improve processes and work more seamlessly across functional lines. In this context lean methods have become a front runner in discussions at drug discovery meetings. What are these methods and are they delivering what is promised, or are we looking at yet another management initiative?
Section Editors: Phil Gribbon, Kelvin Lam
Introduction
There is an ongoing debate whether overall productivity of drug discovery organisations has declined, but it is undis-puted that cost of the drug discovery process has increased dramatically [1–3]. For a few years we have looked at orga-nisational size as the main factor when looking at productiv-ity and cost. Big pharmaceutical companies have been out of favour, small was seen as key factor for success and the Biotech paradigm was celebrated. More research has shown that this answer is too simplistic, most Biotech companies are not generating a return on investment, and productivity is not high by default [4]. Yet it remains a fact that discovering drugs takes too long and costs too much.
In this situation the industrialisation of R&D has been a buzzword [5] and translated into a focus on high-throughput technologies, automation and centralisation as tools for driv-ing economies of scale. Whilst our efforts have driven down cost per data point, they have not decreased overall timelines or cost and led to the realisation that we need to combine operational and scientific excellence, because drugs are not a ‘guaranteed’ output of a highly parallelised, scaled-up process [6].
Within iterative lead optimisation, our strategy has frag-mented the overall task into multiple steps residing in dif-ferent groups and created multiple handover points. Overall workload has increased with scientific concepts focusing on the parallel generation of potency, selectivity and ADME data [7,8]. Data delivery must be timely to have impact on che-mical design and timelines have to be short to generate a fast succession of iterative cycles for compound optimisation. This pressure has fostered a process view and driven a con-ceptual separation of tasks depending on a focus on creativity or process [9,10], although this is not always seen as a welcome classification by scientists. However, a focus on operational excellence is crucial to the overall organisation and can, with ever more complex workflows integrating new scientific concepts and multiple internal and external players, be a source of competitive advantage.
With the view firmly on processes we have ‘discovered’ lean/six sigma, methods that are widely applied across indus-tries, to increase business excellence. Within the last two to three years there is evidence that these methods are applied in pharmaceutical Research and Development [10–15]. Within this manuscript I will discuss the changed focus within molecular screening groups, outline the basic principles of
Editors-in-Chief
Kelvin Lam – Pfizer, Inc., USA
Henk Timmerman – Vrije Universiteit, The Netherlands
HTS Revisited
lean/six sigma methods in the early discovery setting and highlight the difficulties of implementation that may prevent a runaway success.
High-throughput focus?
Only a few authors have tried to assess the overall impact of new technologies on drug discovery [16–18], many more have focused on the impact of single aspects like combina-torial chemistry [19,20], compound storage [21], assay tech-nologies [22] and HTS [23,24]. High throughput and investments into the related hardware have dominated the front pages and driven the view on the ‘success’ of technol-ogies overall and, on the basis of this, analysts have ques-tioned the wisdom of technology investments [25]. High-throughput technologies have been accompanied by pro-mises of transformational change rather than continuous step-by-step improvement. Although there is clearly room for improvement, we should not assess the overall impact of technologies in the light of these misguided promises. An industry driving science and asking for a premium price for innovative products relies on new technologies and concepts. But we need a realistic view of the impact, and of the time-frame for any technology to be integrated as routine tool. And we have to learn from the ‘high-throughput strategy’. Doing something in high numbers cannot be a strategy in its own right and screening 200,000 compounds per day is not in itself a competitive advantage. As a consequence some of these uHTS technologies have disappeared, for example [17] and Aurora uHTS at Pfizer (Pfizer internal, unpublished). Less is more in many cases, but judged on the literature (Fig. 1) it seems that not all organisations or scientists have learnt this important lesson. Instead of screening ten million com-pounds we are investing in the quality of the compound collection and the right assays. Innovative screening con-cepts, iterative screening, fragment-based screening, biophy-sical and structure-based methods are driving highly tailored work streams for individual projects. As Keseru¨ and Makara stated: A universal lead-discovery process devoid of traps and artefacts remains elusive [26].
There are also more subtle developments within molecular screening that have been truly enabling, for example the widespread application of frozen cells [27] that has reduced the burden of traditional just in time cell culture, improving reproducibility, providing back-up for reagent production fail-ures and enabling the outsourcing of routine cell production. Transient technologies, for example such as the use of bacu-lovirus, have reduced the need for stable cell lines in low use applications and testing before committing to further work [28,29]. There is more innovation in the making, but with technologies like high content screening, use of primary cells or label free technologies we need more work to assess the wider impact and should resist the risk to damage valuable technologies by overestimating the likelihood of early success.
We have seen a big increase of in vitro screening, not just driven by HTS but within the follow-up process. Today suc-cess is more difficult due to our increased knowledge and anticipation of problems and higher hurdles introduced by regulators. We produce more screening results to support candidates for further development (Fig. 2) than ever before. In this context the role of technologies is to help putting innovative ideas into practice, testing a hypothesis that pre-viously could not have been tested. It needs investment into the whole process before we can realise benefit from a tech-nology directed at a single process step. It is the ability to flexibly integrate an array of technologies into project work streams that will differentiate successful organisations from also rans [17].
Organisational structures supporting in vitro screening
Organisational structures have evolved in response to the cost, specialised knowledge and support required to execute in vitro screening well (Fig. 3). This is clearly evident in high-throughput screening where we see centralised functions in the majority of larger organisations (and because of the capital investment required several CRO’s offering the service to smaller organisations). The boundaries of these units are blurred: some are focused on full file screening, but others incorporate knowledge-based approaches, characterisation of hits or generation of leads. For lead optimisation, that is the iterative follow-up and candidate characterisation and selec-tion, there are at least three different operating models:
Figure 1. High-throughput strategy. Shown is the number of publications with high throughput in title and abstract for the years 1995–2007 (source Medline).
distributed screening in therapeutic or disease areas, open access approaches where equipment and laboratory space is centrally administered but manned by individual therapeu-tic/disease areas, and a fully centralised model where screen-ing has been centralised into one group. Several support functions are integrated or closely associated with molecular screening operations: reagent provision groups, compound storage and logistics, informatics support and, for organisa-tions relying on large-scale automation and engineering support. It is difficult to draw a line around what is useful; one should rather ask the question which functions need to be closely aligned. For example, it may be useful to have computational chemistry, reagent development or de novo assay development aligned, but these are not essential as core functionality. Outsourcing is now a standard feature in lead optimisation [30] and this has to be aligned with in-house efforts.
The decision about number and location of groups dedi-cated to screening is contingent on several factors. The effectiveness of a group is not linear, but cost (not just monetary but also factors like flexibility, ease of running and decision making) will show an optimum, too many tasks or too big a group will actually decrease the gain. Serving too few projects will make the group inefficient with respect to the utilisation of human and capital assets. In organisations
with several smaller disease areas centralisation into one group can give an advantage, if disease areas are very big it could mandate a team working with just one area, giving greater ownership and collaboration, though not necessarily the most efficient use of key scientific and technical skills that are limited within the organisation.
The core of lean methods
Principles of lean and six sigma methods are well described in the literature and the reader is referred to the wealth of primary and secondary literature as cited in [10–15]. There are a growing number of publications outlining the applica-tion to the Drug Discovery setting, focusing on specific aspects like rapid library synthesis [11], Drug Metabolism and Pharmacokinetics [12], biopharmaceuticals generation [13], in vitro screening [14] and the overall process of iterative lead optimisation [10,15]. Currently we see the main applica-tion to processes which, despite being driven by scientific principles, resemble production environments, although in principle lean methods can be applied in any setting.
It is worthwhile to look at some basic principles as lean (and six sigma) methods are easily misinterpreted. Lean methods will not tell you the right science but aim to execute projects efficiently and with lasting impact, by building processes that are less wasteful. Lean does not try to prevent
Figure 2. Focus on in vitro pharmacology and screening. Exemplified for Pfizer Sandwich the graph shows the overall number of endpoints and the average number of endpoints per candidate (CAN) for further development as expressed of percentage of the 1995 efforts (bars, total Sandwich endpoints; triangles, endpoints per Sandwich CAN).
experiments that give the ‘wrong’ answer as science is about trial and error. We try to prevent experiments that fail to give an answer owing to error (six sigma focus on quality and failure rate) and to prevent experiments giving answers long
after the answer is useful for the project (cycle time focus). Lean methods reconcile the need to execute Drug Discovery projects with utmost flexibility, responding to the emerging data, yet drive cost and timeline reduction in areas like lead optimisation where centralisation and production like pro-cesses are widespread.
Whilst the focus is often on lean, we should look at an integrated concept (Fig. 4). Lean will have dramatic impact on timelines, but for lasting improvement and change we need, at some point, to focus on the underlying process steps reducing variation and failure. Typically, at the research laboratory level, the process is started with a focus on basic stability, establishing useful performance measures and try-ing to organise the laboratory (visual workplace [31,32]). Standardisation is a key method to reduce error rates and elimination of human error caused by uncertainty and lack of training should be the preferred choice before trying to apply more complex six sigma methods [33]. For ongoing contin-uous improvement, structured methods like Kaizen [34] pro-vide a simple tool applicable to even the smallest of problems without big overheads. The cycle of define, measure, analyse, implement and control (DMAIC) underpins Kaizen and six sigma and is easily understood by scientists who are used to analyse problems and focus on numbers.
Fundamental to lean methods is the understanding of our work from a ‘customer’s’ view. Importantly what the custo-mer sees as ‘adding value’ is often very different from our own perception, or the need of a centralised, efficiency driven group. Whilst it may be efficient to group samples from
Figure 3. Organisational structures supporting HTS and lead optimisation. HTS is centralised in most large organisations, with varying boundaries to the follow-up stage, organisational models for lead optimisation and candidate selection vary across the industry (for details see text).
Figure 4. A staged approach for success. A staged approach is driving lasting change with basic stability and continuous improvement underpinning the success of lean transformation (defined as reworking all processes with input from customers and laboratory scientists, with the aim to remove inventory, waiting times and unproductive process steps/working practices (=waste)).
multiple projects for a week to process on large-scale auto-mation, this will clearly add delay to the timelines of the single project. Lean is powerful because we map the current process with all contributing parties (beyond departmental boundaries) highlighting problems with the people doing the work not only taking the managers’ view of what happens. Our workshops within iterative lead optimisation have also included external collaborators for example in the area of outsourced library synthesis and production of frozen cells. These have identified steps in the workflows duplicated in-house (multiple checkpoints), unproductive time through waiting on arrival of samples (which could be fixed by con-ducting electronic work whilst samples are in transit) and misalignments between different departments which have added a whole week to the processing and caused widespread frustration. These improvements have come for free and are possibly by a detailed mapping of what actually happens on paper or in the laboratory (walking the process).
Conclusions
Reading the literature and on the basis of my own experience, the case for lean methods is compelling, yet we have seen single case studies rather than widespread success. In part this is due to the premature claims that methods are widely used in pharmaceutical companies, fostered by a handful of pub-lications and presentations. We seem already to move on to the next topic and initiative and this is a similar pattern observed with the implementation and application of new technologies where the claim of success can precede impact by several years. But there are more substantial hurdles, because the introduction of lean methods requires a cultural change in whole organisations. Any manager can look at their department and even a narrow focus can be truly transformational and set the example for others to follow. This ‘wild fire’ approach, relying on individuals to carry the message further takes a long time and the outcome is by no means certain. At the organisational level lean transforma-tion requires backing from management and colleagues alike. It is too easy to think that changes agreed by a lean workshop will be translated into practice, especially when work across organisational boundaries is required. The challenge of lean methods is the problem of all change management.
A standard phrase in many discussions is ‘But we are not manufacturing cars’ when relating to the Toyota experience. Whilst we could lament scientists not taking a step a back and apply a more abstract view, attempts should be made to package lean into more acceptable delivery for scientists. Scientists are driven by a belief that their way of doing things is best. In this culture of ‘individual creativity’ we should not underestimate the reservations towards an abstract manage-ment approach as illustrated by literature opinions linking the ‘decline’ of pharmaceutical research to the introduction of non-scientists as managers [35].
The challenges for molecular screening and profiling groups are no longer new technologies or scientific concepts, but the swift integration of highly flexible strategies tailored to each project. However, our drive for efficiency and econo-mies of scale has resulted in a fragmented drug discovery process with highly specialised teams, centralised processing steps and project contributions scattered across continents and companies. For further advancement we need to develop processes that can operate in this complex setting without friction and time delay. Lean/six sigma methods are the tools of choice, with no large upfront investment and the ability of scale up from individual teams to whole work streams. The methods can deliver quick results at individual stages, but we should not underestimate the cultural change necessary to integrate lean thinking in every task and team. Given the length of the drug discovery process it is far too early to measure the impact at the macro-level. For the time being we need to concentrate on the success delivered on individual projects and processes. After all every little step helps.
Acknowledgements
I would like to thank Drs Wilma Keighley, Esther Schmid and Rob Spencer for sharing thoughts and ongoing discussion.
References
1 Booth, B. and Zemmel, R. (2004) Prospects for productivity. Nat. Rev. Drug Discov. 3, 451–455
2 Hamilton, M. et al. (2005) Financial anatomy of biomedical research. JAMA 294, 1333–1342
3 Schmid, E.F. and Smith, D.A. (2005) Is declining innovation in the pharmaceutical industry a myth? DDT 10, 1031–1039
4 Pisano, G.P. (2006) Can science be a business? Harv. Bus. Rev. HBR.ORG 5 Handen, J.S. (2002) The industrialisation of drug discovery. DDT 7, 83–85 6 Witty, A. (2008) In Cash-Rich Drugmakers Eye Mergers or Aquisitions,
Weintraub A, Business Weekhttp://biz.yahoo.com/bizwk/081120/ 0845b4107044224586.html?.v=1
7 Caldwell, G.W. et al. (2001) The new pre-preclinical paradigm: compound optimisation in early and late phase drug discovery. Curr. Top. Med. Chem. 1, 353–366
8 Di, L. and Kerns, E.H. (2003) Profiling dug-like properties in discovery research. Curr. Opin. Chem. Biol. 7, 402–408
9 Schmid, E.F. and Smith, D.A. (2002) Should scientific innovation be managed? DDT 7, 941–945
10 Ullman, F. and Boutellier, R. (2008) A case study of lean drug discovery: from project driven research to innovation studios and process factories. DDT 13, 543–549
11 Weller, H.N. et al. (2008) Application of lean manufacturing concepts to drug discovery: rapid analogue library synthesis. J. Comb. Chem. 8, 664– 669
12 Hammond, C. and O’Donnell, C.J. (2008) Lean six sigma – its application to drug discovery. Drug Discov. World Spring, 9–19
13 Jagschies, G. (2008) Operational excellence – the future of biopharmaceutical manufacturing. Inn. Pharm. Technol. 57–60
14 Sewing, A. et al. (2008) Helping science to succeed: improving processes in R&D. DDT 13, 227–233
15 Petrillo, E.W. et al. (2007) Lean thinking for drug discovery – better productivity for pharma. Drug Discov. World Spring, 9–14
16 Schmid, E.F. and Smith, D.A. (2006) R&D technology investments: misguided and expensive or a better way to discover medicines. DDT 11, 775–784
17 Houston, J.G. et al. (2008) Case study: impact of technology investment on lead discovery at Bristol Myers Squibb, 1998–2006. DDT 13, 44–50
18 Amir-Aslani, A. and Negassi, S. (2006) Is technology integration the solution to biotechnology’s low research and development productivity? Technovation 26, 573–582
19 Golebiowski, A. et al. (2001) Lead compounds discovered from libraries. Curr. Opin. Chem. Biol. 5, 273–284
20 Golebiowski, A. et al. (2003) Lead compounds discovered from libraries: Part 2. Curr. Opin. Chem. Biol. 7, 308–325
21 Lane, S.J. (2006) Defining and maintaining a high quality screening collection. DDT 11, 267–272
22 Bender, A. et al. (2008) Which aspects of HTS are empirically correlated with downstream success? Curr. Opin. Drug Discov. Dev. 11,
327–337
23 Fox, S. (2006) High throughput screening: update on practices and processes. J. Biomol. Screen. 11, 864–869
24 Gribbon, P. and Sewing, A. (2005) High throughput discovery: what can we expect from HTS. DDT 10, 17–22
25 Landers, P. (2004) Human element: drug industry’s big push into technology falls short. Wall Street J. In: http://www.wsj.com
26 Keseru¨, G.M. and Makara, G.M. (2006) Hit discovery and hit to lead approaches. DDT 11, 741–747
27 Kunapuli, P. et al. (2005) Application of division arrest technology to cell-based HTS: comparison with frozen and fresh cells. Assay Drug Dev. Technol. 3, 17–26
28 Katso, R.M. et al. (2005) J. Biomol. Screen. 10, 715–724
29 Wang, D.Y. et al. (2008) Development of a high throughput cell-based assay for 11beta-hydroxysteroid dehydrogenase type 1 using BacMam technology. Mol. Biotechnol. 39, 127–134
30 Clark, D.E. (2007) Outsourcing lead optimisation: constant change is here to stay. DDT 12, 6270
31 Hirano, H. (1995) 5 Pillars of the Visual Workplace, Productivity Press, ISBN 1-56327-123-0
32 Ullman, F. (2008) Physical layout of workspace: a driver for productivity in drug discovery. DDT 13, 374–378
33 Tom McCarty. Six Sigma at Motorolawww.motorola.com/ motorolauniversity
34 Kaizen, M.I. (1986) The Key to Japan’s Competitive Success, McGraw-Hill/ Irwin, ISBN 0-07-554332-X
35 Cuatrecasas, P. (2006) Drug discovery in jeopardy. J. Clin. Invest. 116, 2837–2842