Small-molecule libraries for pharmaceutical applications are reviewed.
Continuing with this annual series of comprehensive surveys of combinatorial libraries, the present review covers small-molecule libraries for pharmaceutical applications reported in the literature during 1999. The total number of libraries published in 1999 was 292. There were 85 citations for libraries describing biologically active agents and 207 citations for library constructs without disclosed biological activity. Overall, these numbers are quite similar to those in the previous annual review. There, the first example of an efficacious and orally active compound obtained directly from an optimization library was reported. In addition to new examples of orally bioavailable agents coming from chemical libraries, this year marks another milestone: A 500-member optimization library played a defining role in the identification of a clinical candidate. The effort was reported by Agouron Pharmaceuticals in its structure-based rhinoviral 3C-protease inhibitor program. Achievements such as these are worth noting, as large capital investments have been made in combinatorial chemical technologies. Today, combinatorial sy
 |
 |
| Figure 1. Libraries by major class. |
nthesis pervades many aspects of drug discovery, from lead finding, target validation, and lead optimization to enhancing corporate compound collections.
Including the libraries compiled herein, a total of 975 libraries have been abstracted along with their generic structures in this comprehensive review series, which began in 1992 when the first publications of libraries began to appear in the literature.
An analysis of the data collected in the reviews reveals some interesting statistics and trends in combinatorial chemical research (Figures 1–5). Figure 1A graphically illustrates the number of libraries published between 1992 and 1999 as divided into two broad classifications: chemical libraries for which synthesis and biological assay data are reported (disclosed biological activity), and chemical libraries for which only synthesis—and not disclosure of biological activity—was reported (undisclosed biological activity). The number of reports of biologically active libraries grew at a fairly steady pace. The largest single jump (10-fold) occurred in 1995, with a steep rise also occurring in 1998–1999. The 1998 library number of 74 is nearly equivalent to the combined total of the preceding 6 years. The number of biologically active libraries for 1992 through 1999 was 240. In contrast, the number of reports of library synthesis without disclosed biological activity rose at a much more dramatic pace as the nascent field began to take root. In 1992–1994, only 15 libraries of this type in total had been reported, comparable to the 12 biologically active libraries reported for the same period. Library citations (without biological data) increased by a factor of 3 in 1995 to 43 libraries. In 1996, library publications of this genre more than doubled (2.5 times), held steady for 1997, and then doubled again in 1998 to 247 libraries. Libraries with undisclosed biological activity fell back slightly to 207 libraries in 1999. The total number of libraries in this classification is 735, some 70% more than reports of biologically active libraries. This gap is not too surprising because researchers are anxious to demonstrate new chemical methodologies while safeguarding the structures of active library members. Figure 1B shows the cumulative total of both library classifications, which on balance has increased approximately 1.5 times each year.
The early appeal of combinatorial chemistry was creating large discovery-type libraries through synthesis on solid support. In addition to its perceived synthetic advantages, solid-phase synthesis was the overwhelming choice for library construction between 1992 and 1995 (Figure 2). Some 80% of the libraries produced during this time were generated on solid support. Solution-phase synthesis surged to 50% of the total reports in 1996. This surge was led by advances in the development of new solid-phase reagents, scavenger resins, novel fluorous-based separations, and automated liquid–liquid extractions. Publications of solution-phase library synthesis remained steady at approximately 33% in 1997 and 1998, but receded to 1992–1995 levels (20%) in 1999. The data suggest that solid-phase synthesis continues to hold a dominant position in combinatorial synthesis as more chemistries are redeveloped on this medium.
 |
| Figure 2. Solid- vs solution-phase synthesis for all library constructs (1992–1999). |
|
|
 |
| Figure 3. Library contributions by affiliation (1992–1999). |
|
|
 |
|
Figure 4. Libraries by subclass (1992–1999).
|
|
|
|
 |
|
Figure 5. Discovery, targeted, and optimization libraries (biologically active libraries only).
|
|
Figure 3 indicates the origin of library contributions over the past 8 years, for example, from the laboratories of academia or industry. In the combined years 1992–1997, two-thirds of the contributions were from industrial laboratories, with this number remaining relatively constant in 1998. This past year, library affiliations moved to an industry to academia ratio of 1:1. Overall, pharmaceutical and biotechnology industries appear to be the prevailing players in the game of small-molecule combinatorics, motivated by the goal of increasing drug discovery speed and reducing its costs. The majority of academic publications showcased new synthetic methodologies.
Figure 4 reveals the breakdown of libraries by subclass. Biologically active libraries are designated into one of five subclasses. These include proteolytic enzymes (27%), nonproteolytic enzymes (22%), G-coupled protein receptors (GPCRs, 20%), non-GPCRs (17%), and cytotoxic and anti-infective agents (14%). Within the proteolytic enzyme subclass, serine proteases, namely the trypsin superfamily, were the most screened molecular targets. For GPCRs, opioid receptors appear to be the perennial favorite, not so much as a serious molecular target, but as a convenient demonstration of library utility. Libraries without reported screening data also fall into one of five categories: scaffold derivatizations (27%), acyclic synthesis (19%), monocyclic synthesis (28%), bicyclic and spirocyclic synthesis (22%), and polycyclic and macrocyclic synthesis (4%). A widely used scaffold or template for derivatization is the polyhalogenated heterocycles, such as cyanuric chloride and trichloropyrimidine. Substituted fluoronitroaromatics have been especially versatile reagents for the construction of mono-, bi-, and macrocycles. Many of the classical routes to heterocycles have been reported on solid phase.
Focusing on the 240 biologically active libraries published between 1992 and 1999, one can readily distinguish between discovery, targeted, and optimization libraries (Figure 5). For the purpose of this discussion, discovery libraries are defined as typically large in size (>5000 members) and as having no preconceived notions about which molecular target(s) they may be active against. Targeted libraries are biased in their design—defined as libraries that contain a pharmacophore known to interact with a specific molecular target or family of molecular targets. Optimization libraries are defined as libraries in which a lead exists and an attempt is being made to improve, for example, its potency, selectivity, or pharmacokinetic profile. Accordingly, each of the 240 libraries has been examined and categorized into one of the three groups. Between the early years 1992 to 1997, discovery libraries garnered the highest percentage of citations, at 57%. This was twice the percentage of targeted libraries and four times the reported number of optimization libraries. The number of discovery libraries has fallen rather significantly in the past two years from its 57% high to now the lowest in the group at 21%. In the same two-year period, targeted libraries have moved to the top of the charts, rising from 30% to 54%. Optimization libraries rose from 7% (1992–1997) to approximately 20% (1999), equal to the number of discovery library disclosures. It is tempting to speculate whether this represents a true shift in the way combinatorial chemistry is being valued and applied in drug discovery or is an artifact of industrial research released for external consumption. Anecdotal evidence from discussions at recent conferences and literature commentaries suggest targeted library collections biased toward a specific class or family of molecular targets and “lead explosion” libraries may be preferred over large discovery-type libraries. Large libraries certainly offer unique advantages over smaller focused collections, provided they can be designed with druglike properties and screened efficiently and the actives can be readily identified.
One of the criticisms of combinatorial chemistry that may still slow the acceptance of the technology is that the chemistries generally yield structures that are too peptidelike and contain multiple amide bonds. This is a valid concern due to the known pharmacokinetic liabilities, poor druglike characteristics, and difficulty in optimizing these kinds of compounds. Data derived from the biologically active libraries show that of the libraries reported during the 1992–1997 period, approximately 50% were in fact peptide-based (more than three contiguous amino acid residues). Approximately 70% of the libraries incorporated one or more 
-amino acids, and approximately 85% of the libraries contained one or more amide bonds (data not shown). In the combined years 1998–1999, the number of reported peptide libraries fell by more than half to about 20%, most likely reflecting a bona fide loss of interest in these types of libraries. The use of -amino acids in library construction remains high at approximately 50%, as these synthons represent an excellent source of chiral, low-molecular-weight diversity elements.
Finally, the notion that combinatorial synthesis acting alone will accelerate drug discovery research has not been borne out by experience over this first decade. The ideology of a single universal library as a source of leads against a plethora of molecular targets, purported by some, is not credible. What is evident is that combinatorial synthesis is an important technology among a suite of technologies that can be brought to bear on solving drug discovery problems.
This article is an edited version of the opening section of Roland Dolle’s annotated and illustrated survey published in the American Chemical Society’s Journal of Combinatorial Chemistry 2000, 2 (5), 383–433. The complete survey article, which includes 290 references, 10 tables, and 26 figures, is available on the Web at http://pubs.acs.org/hotartcl/jcchff/2000/cc000055xrev.html.
Roland E. Dolle, Ph.D., is senior director of chemistry at Adolor Corporation (Malvern, PA). Send your comments or questions regarding this article to mdd@acs.org or the Editorial Office by fax at 202-776-8166 or by post at 1155 16th Street, NW; Washington, DC 20036. .
