Competitive pressures and economic realities demand that pharmaceutical developers extract as much value as possible from drug discovery programs. Complicating this imperative are the seemingly endless paradigm shifts within discovery science. Although medicinal chemistry has always been part of the picture, over the last 20 years, the discovery model has evolved from pure med chem to incorporate varying components of computational chemistry, high-throughput methods and rational design. Today's discovery programs are likely to employ all these methods.
While the majority of small-molecule drug discovery efforts share the goal of identifying and promoting active compounds, programs may have any number of starting points. The high-throughput strategy, popular a decade ago and still predominant at many companies, relies on the synthesis or acquisition of large compound libraries and rapidly testing each molecule against appropriately designed in vitro screens. Large, chemically diverse libraries provide the most value in situations where drug target data is sparse, or where ligand classes remain undefined. Large libraries' varying degrees of specificity toward common target classes nevertheless represent a reasonable first-pass attempt to uncover activity, especially when a robust assay is already in place.
Molecular design techniques, which have proliferated of late, seek to reintroduce rational design into library generation. Most major pharmaceutical companies publish and speak extensively on their design efforts. In silico techniques have similarly emerged for virtually screening very large compound libraries. Drug firms' interest in molecular design suggests a small but growing disinclination to pin hopes of discovery success entirely on large, chemically diverse, synthesized compound libraries. The growing popularity of focused libraries supports this view.
Commercially-available focused libraries, whose compounds usually number in the low thousands, are attractive in their specified activity against popular drug target classes such as kinases and GPCRs. These focused libraries are almost never guaranteed to show activity against specific enzymes or receptors. When compounds in these libraries do show activity, they often lack the novel, patentable chemistries essential for commercial viability. An approach that solves the dilemma of achieving novelty and activity together, early in the discovery process, has the potential to deliver an excellent return on investment.
Navigating chemical space
A substantial amount of drug target information exists in the public domain through patents, journals, receptor and gene databases, ligand and crystal data. Theoretically, anyone can design a quantitative structure-activity relationship (QSAR), pharmacophore, or receptor-based model based on these data. Design packages such as SYBYL (Tripos), Cerius2 (Accelrys), and Phase (Schrodinger) greatly improve the odds of successfully mining non-proprietary data for activity. The difficulty arises in applying resulting models to validated chemistry.
Perhaps the most daunting hurdle to success under this paradigm is the huge size of chemical space. How can chemists search this space efficiently, using appropriate, efficient, goal-driven criteria? And as success beckons, how does one avoid the significant swaths of chemical space already staked out by competitors?
Clearly, the challenge for modern discovery and development companies is to generate or acquire focused libraries whose molecules are chemically unique, active in specific target areas, flexible enough to be modified for specific sub-families or individual targets, and which appropriately balance the risks and costs of early-stage discovery against the likelihood of eventual success.
High-throughput methods
High-throughput screening (HTS) is inherently more expensive, on a per-compound basis, than library synthesis. Developing a robust, high-throughput assay panel and subsequent screening of 50,000 compounds typically represents an investment of about $500,000. That compound library may be acquired from a variety of vendors, for roughly one dollar per compound. It is common in today's high-throughput screening environment to run a primary screen through a million-entry compound library in a matter of days. Although screening may indeed be rapid, assay development involves weeks of planning, engineering and fine-tuning to achieve acceptable speed, robustness and cost effectiveness. Post-screen data analysis adds significantly to this effort.
Consequently, discovery organizations employ high-throughput methods more selectively than in the past, and are constantly seeking to increase the number of hits per screen through the use of design software. The most successful efforts link molecular design with synthesis and ADMET. Otherwise, the vast majority of hits will fail due to intractable chemistry, poor pharmacokinetics, unfavorable toxicology or some combination of the three.
Focused libraries are generally regarded as the most efficient work-around for avoiding HTS and its necessary antecedent, parallel synthesis (or library acquisition). When designed correctly, small focused compound libraries exhibit a higher probability for activity against specific targets. At the same time, they save discovery organizations the time and expense of implementing large HTS programs.
Any drug discovery effort must weigh the inevitable costs of large library screening and acquisition versus the potential benefits of lower-throughput screening of more focused libraries.
No-man's land
Although active molecules abound, biologically relevant chemical space is increasingly becoming a no-man's land. Because most of the low-lying fruit has already been harvested, active molecules are more likely than not to belong to patent territory that has already been staked out. Drug developers infamously patent huge numbers of compounds in the chemical space surrounding active molecules. Anyone who reads a pharmaceutical industry patent will note the blanket IP protection enjoyed by successful and unsuccessful molecules alike. Hence, the major challenge for medicinal chemists has shifted from identifying activity, to locating structures and scaffolds within chemical space which are novel and amenable to improvement through follow-on synthesis. To deliver value to discovery efforts, focused compound libraries must reflect these realities.
Scaffold-hopping represents a serviceable first-pass attempt at achieving novelty and activity together. Through this technique, discovery scientists identify novel, patentable structures from known, active chemical backbones. Medicinal chemists have used intuition- and expertise-driven scaffold-hopping for decades. The recent dearth in approval of new chemical entities suggests that straightforward scaffold-hopping, even when aided by software products, has not been successful. A principal reason for this, we believe, is the over population of chemical space by already-patented structures, and the inability of most commercial scaffold-hopping applications to search chemical space rapidly and efficiently.
Scaffold-hopping software is available in several formats. Pharmacophore models, which search molecules by shape and pharmacophore features, are examples of search-limited scaffold hopping. Pharmacophore models can sift through up to several tens of thousands of compounds per week, which by today's standards is quite slow. Also, in addition to sluggishness in searching virtual libraries, most scaffold-hopping methodologies lack innate appreciation of chemical synthesis, a trait that would allow evaluation of structures for synthetic feasibility.
A robust, discovery-worthy scaffold-hopping program should therefore search 10 to 100 times as rapidly as pharmacophore models while providing chemical intelligence.
Conventional scaffold-hopping software also lacks adequate ADMET support. At minimum, the software should recognize and reject pharmacophores known to raise ADMET issues. Structures such as four fused aromatic rings, Michael acceptors, and activated halogens possess well-documented toxicity, while nitro groups are easily metabolized to nitroso groups. Best-in-class scaffold-hopping applications will also flag substituents that are likely to result in poor solubility.
Spreading risk
Acquiring focused libraries can help discovery organizations balance compound activity, selectivity, manufacturability, and IP considerations against time and costs. For example, Tripos' LeadDiscovery programs covering GPCRs and kinases focus on providing a low-risk entry into active, proprietary chemical space for early-stage discovery companies. The combination of advanced scaffold-hopping technology (ChemSpace) with traditional library design tools in LeadDiscovery results in extremely rapid searches of proprietary, druggable chemical space. Subsequently, synthesized compounds are then screened as proof of mechanism in selected target areas.
No compound vendor can promise that any particular library will provide hits against specific targets. Where most library acquisition relationships impose the preponderance of risk on the discovery organization, innovative approaches to compound library acquisition and use can, in some situations, distribute this uncertainty more equitably between vendor and customer.
One approach involves screening compounds blinded, without knowledge of the compounds' chemical structures. Apart from a nominal handling fee, the customer is not obligated to pay for inactive compounds, while the vendor's IP remains protected and available to other discovery organizations.
To give discovery organizations confidence in this approach, they must be sure that libraries belong to novel scaffold families that are either patented or are the subject of a patent application. Every scaffold must be capable of undergoing chemical modification to design in desirable qualities like activity and selectivity, while designing out toxicity. Depending on their results and individual needs, customers should have the option to purchase promising scaffolds outright and develop them on their own, or to hire the vendor to manage subsequent medicinal chemistry manipulation and deliver finished compounds. In either case, second-generation molecules would typically be of sufficient quality to entering more demanding assays or even animal models.
Industry business models to acquire focused compound libraries are part of a growing trend in pharmaceutical discovery, development and even manufacturing towards managing and assessing risk appropriately among various stakeholders. As part of that evolution, discovery organizations will increasingly seek relationships that provide value while minimizing up-front exposure to the great unknowns in achieving activity and novelty together.
After Dr. Mark Warne completed a chemistry doctorate from Bristol University, he joined Tripos Discovery Research as a computational chemist, and then moved to group leader. In 2003, Dr. Warne was promoted to collaborations coordinator, where he provides business support and establishes scientific relationships with biotech and pharmaceutical partners.
