Redefining Drug Discovery Informatics
As the pharmaceutical industry changes in response to unprecedented challenges, information technology supporting the drug discovery process must evolve and be redefined in order to improve research productivity and success rates.
Improvements to traditional computational chemistry applications will continue to provide incremental benefits, but drug discoverys productivity imperative can only be met from the more widespread use of vital decision support tools and through discovery informatics platforms that foster highly collaborative discovery processes.
For decades, the pharmaceutical industry’s highly profitable blockbuster business model has shielded drug companies from a sense of urgency to reinvent themselves and the processes by which they discover drugs. However, the industry is now facing significant productivity challenges – specifically how to efficiently generate enough pipeline-stage products to replace fading blockbusters, and to deliver on the promise of genomics. As a result, the pharmaceutical industry is on the brink of significant change, requiring the reinvention of traditional discovery processes and the redefinition of the information technologies required to support these processes.
Drug discovery’s productivity imperative has been brewing for years. Loss of patent protection for blockbuster drugs, an over-reliance on highthroughput discovery strategies, and ongoing industry consolidation through mergers and acquisitions are in large part to blame for the current state of affairs, in which rising R&D expenses and downward pricing pressures are threatening to pull the industry apart.
The blockbuster business model, which requires a steady stream of new products to justify high sales and marketing expenses, appears outdated. Thanks to genomics and proteomics, we are beginning to think of drug safety and efficacy in terms of differential gene expression rather than broad disease categories such as ‘breast cancer’. Drug discovery as currently practised cannot respond satisfactorily to the need, say, for 20 new blood pressure medicines when just 20 or so New Chemical Entity (NCE) approvals per year are the industry’s norm.
Pharmacogenomic approaches to individual therapeutic categories such as oncology may, in fact, require hundreds of new drugs. Most likely, next generation drug therapies will employ novel mechanisms of action and new structures. Clearly, the industry requires an approach that weaves vital information, IT platforms and processes together; discovery informatics must rise to meet these new challenges.
Innovation and outsourcing
Before embarking on possible solutions, the industry must re-examine traditional business models and rethink, as some have suggested, the role big pharma plays in drug discovery.
The growth rate of pharmaceutical R&D investment has been steadily increasing – the 13% increase between 2002 and 2003 in R&D expenditures by US-based pharmaceutical companies to $35.5 billion is by any standard a huge number. However, the industry’s current R&D investment is neither optimal nor sustainable, considering the fact that the top 50 pharma companies have produced on average less than one NCE since 1998 (1).
Furthermore, the industry that was once highly respected and celebrated for its contributions to the world now faces ongoing backlash from society for high drug prices, despite escalating R&D costs, and for perceived drug safety problems, regardless of the benefits that millions of people realise every day from pharmaceutical products. An increasing regulatory burden and the lack of international regulatory harmony only adds to the challenges and frustrations facing pharmaceutical executives.
The pharmaceutical industry’s singular responsibility to provide safe, effective treatments engenders a good deal more uncertainty and risk than, say, developing a new aircraft. The most attractive new drugs, as measured by health benefits and stockholder returns, tend to be new molecular entities possessing novel modes of action rather than. so called ‘me-too’ or ‘lifecycle’ drugs. Since these products are based on ground-breaking science, they significantly raise the level of risk of bringing them to market. Pharmaceutical discovery is, in fact, all about risk and uncertainty, and increasingly, how to minimise those factors while efficiently getting products to market and providing value to stockholders and healthcare consumers.
Some experts now believe that big pharma should supplement the drug discovery process by outsourcing to companies whose corporate cultures and financing structures are more compatible with the risks of early-stage research, meanwhile leveraging their own key competencies such as partnering, marketing, manufacturing and managing product lifecycles. According to a 2004 report by Kalorama Information, US pharmaceutical firms spent nearly $2 billion in 2003 on outsourcing discovery. By 2006 the figure is expected to reach $6 billion (2).
Management consulting and technology services company Accenture has described a drug development paradigm that relies in part on “leveraged alliances” for new molecules. Financial models also suggest that even when pharma pays a significant premium to acquire promising candidate compounds or clinical-stage molecules, the economics favour the in-licensing of late-stage compounds over the pursuit of internal, and risky, early-stage discovery.
As this occurs, both discovery contractors and inhouse discovery teams will find themselves subject to cost and productivity accountability normally associated with manufacturing or commodity product development. Within this ‘perform-or-perish’ environment, truly significant gains in productivity will arise from a mix of discovery process improvements and new information technologies to support these new processes.
The productivity, innovation and efficiency gains the industry requires will come not solely from new point discovery informatics applications working independently, but through software platforms that enable greater accessibility to information, more rapid and autonomous decision- making, and collaboration among discovery teams. As the pharmaceutical industry continues to undergo significant change, discovery informatics is also being redefined to satisfy the new demands of more innovative science, improved laboratory efficiency, and, perhaps most importantly, research team collaboration.
In the beginning: computational chemistry
The quest for greater efficiencies in drug discovery is not new. The term ‘rational drug design’ first came into widespread use during the 1980s, followed by high-throughput and ultra-high throughput synthesis and screening methods. Aided by discovery informatics, such as tools to quantitate the diversity of large compound collections, discovery organisations acquired and tested millions of compounds.
However, these companies quickly learned that brute-force techniques were not effective at generating novel, pharmacologically active structures. While the combinatorial chemistry craze did result in several important innovations, such as improvements to synthesis, purification and QC technologies, high throughput methods generally failed, and the industry lost ground by focusing on compound quantity rather than quality. Nonetheless, discovery informatics applications have made a positive impact on the industry thus far and most pharmaceutical companies have employed computational scientists dedicated to using such tools for a decade or more.
Until recently, discovery informatics was traditionally limited to a set of scientific tools reserved for these computational scientists who provide highly specialised modelling and decision support assistance for drug discovery project teams. These scientists, numbering roughly 3,500 worldwide, represent a tight-knit community of specialists whose work in the computer lab is usually a step removed from the work of the medicinal chemists that are busy making compounds in the ‘wet lab’. Therefore, the term ‘discovery informatics’ has traditionally been referred to synonymously with ‘computational informatics’ or ‘computational chemistry’.
Despite the widespread use of traditional computational informatics tools, drug discovery’s evolution toward a growing number of more complicated biological targets, rising expectations from executive management for more accountability for project decisions and shifts to discovery outsourcing cannot occur without an accompanying redefinition of discovery informatics. This redefinition is already occurring, not only with improvements to the underlying scientific algorithms used by computational scientists, but even more so through the adoption of new decision support tools for the broader universe of research scientists. Exciting new technologies are now becoming available to promote collaborative discovery within project teams and between organisations.
Toward redefining discovery informatics
The benefits a discovery company has derived from using discovery informatics software have traditionally been limited by the rate, frequency and effectiveness of the existing communication between the computational scientists (who use specialised software to generate information about what compounds should be synthesised and tested next) and the laboratory scientists (who use the information generated by the software to assist with their decisions in the lab about which compounds to synthesise and test next).
Enabling laboratory chemists, for example, to quickly generate decision-aiding information themselves avoids the often-lengthy delays that occur when busy chemists must wait days or even weeks for results created by equally busy computational chemists who are typically assisting well over a dozen laboratory chemists each.
The substantial corporate cost of having laboratory chemists wait for important decision-aiding information is small compared to the costs of forcing such chemists to make decisions before the software-generated decision support information is available. That cost, in turn, is minimal when compared to the costs of allowing computational chemists to investing time generating such decision- aiding information for decisions that already have been made.
Today’s discovery informatics applications are therefore providing better decision-making tools and greater autonomy to more researchers, including laboratory scientists. The stark reality of sparse pipelines has jolted discovery organisations into recognising that the medicinal chemist’s craft is irreplaceable, and cannot simply be replaced by in silico methods alone. Nor can the industry continue to struggle with poor communication and lack of information-sharing among project team members.
Medicinal chemists themselves are increasingly using new discovery informatics applications to conduct their own preliminary analysis of ligand-receptor interactions, to design small parallel libraries, and to communicate their ideas with colleagues in biology and computational chemistry. This group of ‘front-line’ scientific applications, part of a category of software generally referred to as ‘laboratory informatics’, is a critical component in the redefinition of discovery informatics, and is increasingly being used by many laboratory scientists, including laboratory chemists and biologists.
One example of a laboratory informatics application is the electronic lab notebook, which is quickly replacing paper lab notebooks for the improved storage, searching and sharing of digital discovery information. Other laboratory informatics applications include ‘predictive chemistry’ software, such as tools for virtual library design and structure-activity relationship data mining that are enabling more rapid and autonomous decision making directly in the wet lab.
While advances in the fundamental algorithms employed by computational informatics are leading to improvements in scientific innovation and laboratory informatics is introducing greater autonomy and confident decision making in the lab, a third component of the new discovery informatics – ‘operational informatics’ – is improving the efficiency with which the industry operates its discovery laboratories.
Operational informatics includes solutions for running an efficient and agile laboratory. These technologies include chemical registration systems, chemical inventory systems, reagent order management systems, and other IT systems that record the flow of chemicals in, out of, and throughout the drug discovery lab. Traditional chemical registration systems, for example, have been used to capture, store and secure the very intellectual property upon which the pharmaceutical company is based – experimental information about millions of important chemical compounds, each of which may possess the properties required to become one of the company’s next profit-producing and life changing new drugs.
However, in the face of mounting industry pressures to transform discovery processes and the IT systems that support them, even traditional enterprise chemical registration systems are evolving beyond simply capturing experimental ‘data’. The real value of an informatics system lies in its ability to record the context in which laboratory experiments are performed, leading to real and comprehensive ‘knowledge’ sharing among researchers.
In this context, chemical registration systems are being redesigned and recast as vital knowledge bases to store computational models, virtual libraries and discovery workflows, none of which are stored in today’s leading registration systems. Registration systems are also being overhauled to more accurately secure and represent a company’s chemical intellectual property. The fact that modern registration systems do not clearly discern between or link in silico experimental results and in vitro experimental results is surprising and leads to problems that could potentially cost a company its next blockbuster compound.
Like other components of discovery informatics, operational informatics systems will also evolve beyond providing improvements to point applications for individual researchers. Major boosts in productivity will accrue from operational software that addresses the range of laboratory needs holistically, providing all team members with the benefits of information and knowledge generated by their colleagues.
On-demand access to important operational data, such as the location of bar-coded reagent vials or the availability of certain chemicals in the company storehouse, as well as the ability to access the results from previously performed experiments can shave weeks off individual scientific effort, and months of team effort. Clearly, improvements to operational informatics technologies, especially when combined with computational and laboratory informatics tools, stand to contribute significantly to the industry’s current productivity problems and to the redefinition of discovery informatics.
High-performance computing and web services: catalysts for change
A discussion of the transformations taking place within discovery informatics would not be complete without briefly mentioning two key advances in scientific computing that are contributing to these changes. High-performance computing solutions coupled with flexible, service- oriented software architectures are quickly becoming the standard for modern discovery informatics environments. Large and small discovery organisations are using these platform technologies as foundations upon which to build and deploy computational, laboratory, and operational informatics solutions.
Powerful, inexpensive computing has been, and will continue to be, a catalyst for transforming modern discovery informatics. Migration from expensive mainframe computers and large, shared clusters to small workgroup clusters, ‘personal supercomputers’, and powerful desktops and laptops is opening up the full potential of next generation discovery informatics to a broader range of scientists than ever before. The burden now lies on software developers to create scientific software products that earn high marks for both ‘learnability’ (how fast a user can learn how to use the software) and usability (how easy the software is to use once it has been learned) in order to exploit this inexpensive computing power most effectively.
Software developers will also be challenged to exploit Moore’s law not only to achieve faster results from today’s discovery informatics programs, but also to enable the development of novel approaches to existing discovery challenges. Thus, next-generation discovery informatics will also introduce ‘next-generation science’ that will be at the heart of computational, laboratory and operational informatics applications leading them not simply to ‘go faster’, but to approach problems in better, more effective ways than previously thought practical.
As discovery informatics is redefined, advances in scientific innovation alone are not sufficient. The industry standard for discovery informatics software architecture in pharmaceutical R&D is also changing. The workstation-centric platform that has served the industry well for many years is still important. But now, more loosely-coupled, server-centric architectures featuring software deployed as flexible web services that can adapt to each company’s unique discovery workflows are critical for improving the agility and productivity of cross-functional project teams. Currently, traditional software is being redeployed as flexible discovery informatics web services, creating opportunities to broaden access to useful decision support tools and other software to a much larger base of researchers.
The redefinition of discovery informatics, from computational tools reserved for small groups of specialised scientists, to laboratory decision support tools and knowledge bases accessible to all members of the discovery project team is already under way. Many early-adopters are already benefiting from these new applications, while other pharmaceutical companies are now busy organising cross-functional working groups and executive steering committees to evaluate how such applications can best be applied to their unique discovery processes. Given the problems facing the industry today, it is clear that something must change. Companies must find new ways to realise the return on R&D investment they require in order to improve discovery process performance.
For now, the process improvement effort and application of modern discovery informatics tools continue to focus on intra-company decision making and collaboration improvements between fellow researchers and between project teams within the enterprise. In the near future, however, as the pressure to rethink the status quo in drug discovery increases in the face of unprecedented challenges to traditional discovery methods, and as the industry reinvents itself to focus on core competencies and supplement via outsourcing those areas that are viewed as either non-core or too risky, tools that foster even more efficient collaboration will take on a more central role.
Discovery informatics will continue to evolve to facilitate highly collaborative discovery processes, both internally and with external collaborators across the globe. Project team members, including internal scientists from external collaborators, will securely and efficiently share discovery data including computational models, in silico and in vivo experimental results, and drug safety information, leading to a sea change in the productivity of the entire industry.
It is unclear where the industry will be five years from now. But what is clear is that disruptive change is under way and the discovery informatics technologies that made their debut some 20 years ago were only the beginning for what will be required when the dust settles after this next wave of change in the pharmaceutical industry. DDW
Until its purchase by Tripos, Inc, Bryan Koontz was Chief Executive Officer of Optive Research, Inc, a molecular discovery software company he co-founded in 2002 along with Professor Robert S. Pearlman at the University of Texas. Prior to Optive Research, Mr Koontz served in a variety of executive roles, including VP Marketing and VP Business Development, at Contextual, Inc, an enterprise software company co-founded by Mr Koontz and backed by Dell Ventures, Austin Ventures and TL Ventures. Mr Koontz also held several management positions, including Director of Business Development and Director of Product Management at Trilogy Software, Inc from 1997 to 1999. Mr Koontz holds a BS with honours in Mechanical Engineering and a minor in Engineering Mechanics from The Pennsylvania State University and graduated summa cum laude with a MS in Mechanical Engineering with a focus in control system theory from The Massachusetts Institute of Technology. He is currently Senior Vice-President and General Manager of Discovery Informatics at Tripos, Inc.
1 Deloitte and Med Ad News. Challenges and Opportunities Converge in Today’s Pharmaceutical Industry. February 2005.
2 Kalorama Information. Outsourcing in Drug Development:The Contract Research Market from Preclinical to Phase III. January 2004.