Drug candidates traditionally have been synthesized one at a time, a time-consuming and labor-intensive process. But many researchers in academia, government, biotechnology firms, and drug companies increasingly are turning to combinatorial chemistry - a strategy for creating new drugs that, it is hoped, will speed the drug discovery process significantly.
The idea of combinatorial chemistry is to make a large number of chemical variants all at one time; to test them for bioactivity, binding with a target, or other desired properties; and then to isolate and identify the most promising compounds for further development.
The success of combinatorial chemistry is still uncertain. No drugs discovered combinatorially have been approved for marketing, although several are currently in development. But many researchers believe the technique will prove to be an efficient and cost-effective tool for identifying new medicines.
In combinatorial chemistry experiments, chemical libraries (large collections of compounds of varied structure) are produced by sequentially linking different molecular building blocks, or by adding substituent "decorations" to a core structure such as a polycyclic compound. Libraries may consist of molecules free in solution, linked to solid particles or beads, or even arrayed on surfaces of modified microorganisms.
Combinatorial chemistry initially focused on the synthesis of very large libraries of biological oligomers such as peptides and oligonucleotides. But drug developers generally prefer to focus on small organic molecules with molecular weights of about 500 daltons or less - the class of compounds from which most successful drugs have traditionally emerged. So combinatorial chemistry researchers are concentrating on small organic compounds as well.
Drug discovery is the primary goal of most combinatorial chemistry research, but combinatorial methods also have potential applications for development of advanced materials and catalysts.
One of the challenges of combinatorial chemistry is the difficulty of identifying "hits" (active compounds) present at vanishingly low concentrations in complex combinatorial libraries. To address this problem, ingenious encoding schemes have been developed. Two groups have independently developed the latest concept in this field - radiofrequency encoding, in which information about library compounds is stored on microchips.
Instrumentation systems to help speed combinatorial chemistry experiments have been developed in-house at a number of biotechnology and pharmaceutical companies. And several combinatorial automation systems are available commercially or undergoing intensive development.
Combinatorial chemistry has come a long way in just a few years, but further advances are needed and new applications are anticipated. Directions in which the field is headed range from combining combinatorial chemistry with computational drug-design strategies to the use of combinatorial molecular recognition for studies of protein function.
Combinatorial libraries are created in the laboratory by one of two methods - split synthesis or parallel synthesis. In split synthesis, compounds are assembled on the surfaces of microparticles or beads.In each step, beads from previous steps are partitioned into several groups and a new building block is added. The different groups of beads are then recombined and separated once again to form new groups. The next building block is added, and the process continues until the desired combinatorial library has been assembled.
Before split synthesis was developed, explains chemistry professor Kim D. Janda of Scripps Research Institute, La Jolla, Calif., "people created diversity using mixtures of compounds. In a coupling step, you would add, let's say, reagents A, B, and C in one pot, and A, B, and C would all compete to become integrated at the same site. But in doing that you can have problems with kinetics. One reaction may be faster than another and you may not get equal distribution of the three components."
Split synthesis "got away from all that," says Janda. "You could create diversity using separate reactions, so the components would have an equal chance to add in to a site, and then by mixing compounds together again you got the diversity you needed."
Libraries resulting from split synthesis are characterized by the phrase "one bead, one compound." Each bead in the library holds multiple copies of a single library member. Split synthesis greatly simplifies the isolation and identification of active agents because beads (and implicitly individual library members) are large enough to be observed visually and separated mechanically.
Combinatorial libraries can also be made by parallel synthesis, in which different compounds are synthesized in separate vessels (without remixing), often in an automated fashion. Unlike split synthesis, which requires a solid support, parallel synthesis can be done either on a solid support or in solution.
A commonly used format for parallel synthesis is the 96-well microtiter plate. Robotics instrumentation can be used to add different reagents to separate wells of a microtiter plate in a predefined manner to produce combinatorial libraries. Hits from the library can then be identified by well location.
Split synthesis is used to produce small quantities of a relatively large number of compounds, whereas parallel synthesis yields larger quantities of a relatively small number of compounds. And split synthesis requires that assays be performed on pools of compounds, whereas assays on individual compounds can be run on libraries created by parallel synthesis. While slower, testing individual compounds is sometimes advantageous because serious interferences and complications can arise when multiple compounds are tested simultaneously.
A special case of parallel synthesis is spatially addressable synthesis, pioneered by researchers at Affymax Research Institute, Palo Alto, Calif. In this technique, libraries are synthesized in arrays on microchips, and all the compounds on a chip are assayed simultaneously for binding or activity. Hits can then be identified by the piece of real estate they occupy on the chip. Using a chip-making technique called photolithography, Affymax researchers have generated arrays of more than 65,000 compounds on chips about 1 sq cm in area.
Bioactive combinatorial compounds synthesized by split synthesis can also be identified by deconvolution, a technique in which each variable position in a compound library is tested to find the building block that makes the strongest contribution to activity at that site.
Solid-phase and solution-phase combinatorial synthesis each have their advantages and disadvantages. Solid-phase synthesis permits use of excesses of reagents to drive reactions to completion, since excess reagents can be washed away from beads very easily afterward. However, solution-phase synthesis is more versatile because many organic solution-based reactions have not been adapted for solid-phase work.
Janda and coworkers at Scripps recently developed a liquid-phase synthesis procedure that combines some of the advantages of solution-phase and solid-phase synthesis [Proc. Natl. Acad. Sci. USA, 92, 6419 (1995)]. The procedure involves use of polyethylene glycol monomethyl ether in place of solid-phase beads as a foundation for combinatorial assembly. The polymer is soluble in a variety of aqueous and organic solvents, making it possible to use solution-phase combinatorial synthesis. But the polymer can be precipitated out of solution by crystallization at each stage of the combinatorial process to facilitate purifications.
Combinatorial chemistry began with the synthesis of large libraries of biopolymers such as peptides and oligonucleotides. In some cases, these were created on surfaces of genetically modified microorganisms, such as bacteriophage particles, by inserting combinatorial DNA oligomers into genes that encode cell-surface proteins.
However, peptides and oligonucleotides are problematic for drug development because their oral bioavailability is poor and they are degraded rapidly by enzymes. Hence, the focus of combinatorial research has shifted in recent years to libraries of nonpolymeric small molecules having molecular weights of about 500 daltons or less.
In a pioneering study, chemistry professor Jonathan A. Ellman and coworkers at the University of California, Berkeley, synthesized the first such library by creating variants of benzodiazepines, a class of compounds that has been a fertile source of successful drugs [J. Am. Chem. Soc., 114, 10997 (1992)]. Since then, researchers have found ways to synthesize combinatorial libraries based on many other classes of small organic compounds.
A recent example is work by Mark A. Gallop, director of combinatorial chemistry, and coworkers at Affymax. They used a cycloaddition reaction to prepare a small-molecule combinatorial library of about 500 mercaptoacyl prolines [J. Am. Chem. Soc., 117, 7029 (1995)]. By screening this library, they identified an unusually potent inhibitor of angiotensin-converting enzyme (ACE). ACE inhibitors are used as treatments for hypertension and heart disease.
And the group of Stephen W. Kaldor, head of combinatorial chemistry research at Eli Lilly & Co., Indianapolis, in collaboration with scientists in Lilly's central nervous system (CNS) group, has used combinatorial chemistry to identify an orally active CNS agent by combinatorial optimization of an existing lead. The low molecular weight nonoligomeric drug candidate entered clinical trials in November. "This is one of the first small-molecule combinatorial compounds to go into humans," says Kaldor.
A major challenge of small-molecule combinatorial chemistry has been to adapt conventional solution-phase organic reactions to reactions on solid-phase particles. Ellman says one of his group's efforts "has been to expand the kind of chemistry that can be performed on solid supports in a simultaneous synthesis format - in particular, carrying out different types of carbon-carbon bond-forming reactions. For example, we've developed general enolate alkylation conditions where side reactions that can be a problem in solution don't occur."
Paralleling the increasing use of small-molecule libraries is a trend toward assaying libraries having smaller numbers of components. "People seem to be much more comfortable working with smaller mixtures - probably a hundred components or less in a mixture, rather than the mixtures of 105 and 106 compounds per pool that we saw in the early experiments," says Ronald N. Zuckermann, associate director of bioorganic chemistry at Chiron Corp., Emeryville, Calif. "The lower the number of compounds, the more confidence you can have in the biological data" because artifacts arise more readily in the screening of large pools of compounds.
Ellman agrees that "people have gotten away from screening really large mixtures of compounds. They either want to screen them individually or in smaller pools of under 100 compounds. It's easier to extract out binding data in that format."
Carbohydrates have lagged behind other types of compounds in combinatorial library development because of the complexity of oligosaccharide chemistry, but carbohydrate libraries are now beginning to appear. For example, Ole Hindsgaul and coworkers at the department of chemistry of the University of Alberta, Edmonton, in collaboration with researchers at the University of Georgia, Athens, and Ciba Central Research Laboratories, Basel, Switzerland, have developed a "random glycosylation" strategy for making oligosaccharide libraries in solution [Angew. Chem. Int. Ed. Engl., 34, 2720 (1995)]. They produced a library of all 18 possible fucosylated trisaccharides from disaccharide precursors.
And at a recent meeting, chemistry professor Daniel E. Kahne of Princeton University reported construction of the first solid-phase carbohydrate library, using chemistry for solid-phase synthesis of oligosaccharides developed earlier by his group. This technique has been licensed to Transcell Technologies, Monmouth Junction, N.J. In preliminary work, compounds isolated from one carbohydrate library have been shown to bind a carbohydrate-binding protein with greater affinity than the protein's natural ligand. "Carbohydrates play a central role in some very important biological processes, so having access to libraries of these compounds is critical," says Kahne.
Another type of oligomer being pursued combinatorially is peptoids, peptide analogs that are not recognized by peptide-cleaving enzymes. Chiron researchers recently discovered a candidate urokinase receptor antagonist from a peptoid library, and the compound is currently in preclinical studies as a potential anticancer agent.
"One of the primary advantages of peptoids is their synthetic accessibility," says Zuckermann. "They are efficiently synthesized by the submonomer method, which uses primary amines and bromoacetic acid as starting materials - both very cheap, and there are literally thousands of amines readily available. The combination of this chemistry with robotic synthesis has led to a truly high throughput synthesis facility."
Chiron's identification of nanomolar peptoids that bind to transmembrane receptors [J. Med. Chem., 37, 2678 (1994)] "was the first example of the discovery of potent ligands to pharmaceutically relevant receptors from a combinatorial library of nonpeptides or nonnucleic acids - that is, synthetic compounds," Zuckermann adds. "I believe that this work helped inspire others to continue to move away from peptides and further toward small molecules."
Combinatorial chemistry can also be extended entirely beyond the realm of organic chemistry. For example, physicist Xiao-Dong Xiang of Lawrence Berkeley National Laboratory, chemistry professor Peter G. Schultz of UC Berkeley, and coworkers recently devised a combinatorial strategy for finding advanced materials with novel chemical or physical properties - extending "the combinatorial approach from biological and organic molecules to the remainder of the periodic table," as they put it [Science, 268, 1738 (1995)].
Xiang, Schultz, and coworkers used thin-film deposition and physical masking techniques to synthesize libraries of solid-state materials. The properties of the resulting materials were then evaluated to identify promising candidates for further development.
In spatially addressable combinatorial synthesis, active compounds can be identified by location. But in other forms of combinatorial chemistry, identifying hits is not so easy because there's often too little of each compound present for characterization with traditional analytical chemistry techniques.
Hence, many researchers now use some form of tagging or encoding to label compounds in large combinatorial libraries. The first such encoding scheme was proposed in 1992 by Scripps President Richard A. Lerner and molecular biologist Sydney Brenner at the institute. They suggested that a combinatorial library could be encoded with oligonucleotides synthesized in parallel with library compounds and linked to each one. Amplification or decoding of the attached oligonucleotide would serve to identify the small molecule bound to each bead.
This idea was independently arrived at and reduced to practice by scientists at Affymax. Later on, researchers at Chiron and at Selectide, Tucson, Ariz., developed similar techniques in which peptides instead of oligonucleotides were used as the sequenceable encoding oligomers.
In 1993, chemistry professor W. Clark Still and coworkers at Columbia University developed a second major type of encoding scheme, in which chromatographically resolvable organic tags were used as encoding elements for bead-based combinatorial libraries. Still devised the technique in response to concerns about the tendency of DNA and peptide tags to break down under the often very rough conditions of organic synthesis.
In Still's technique, inert halogenated aromatic compounds are used to encode the chemical reaction history experienced by each bead. These tags are identified by capillary gas chromatography to reveal the identity of active compounds in the library. Kahne, who used this type of encoding to construct his combinatorial carbohydrate library, says the method "is as good as it gets for identifying hits - a very simple solution to a very important problem."
The most recent development in encoding technology involves the use of radiofrequency tags. Chemistry professor K. C. Nicolaou at Scripps and the University of California, San Diego, together with senior chemist Xiao-Yi Xiao, President and Chief Executive Officer Michael P. Nova, and their coworkers at IRORI Quantum Microchemistry, La Jolla, Calif., developed a technique in which memory devices are associated or coated directly with derivatized polymer during combinatorial synthesis [Angew. Chem. Int. Ed. Engl., 34, 2289 (1995)]. The chips encode relevant information about the synthetic pathway - including not only reagents used, but also reaction conditions such as temperature and pH. The device can then "report" this information to a receiver via radiofrequency transmission.
"We're putting a manual system to do this type of radiofrequency combinatorial chemistry out on the market in March," says Nova. The system will include radiofrequency memory devices in MicroKans, tiny spherical capsules with porous walls that also enclose polymer beads for combinatorial synthesis.
A related technique was developed independently by synthetic chemist Edmund J. Moran and coworkers at Ontogen Corp., Carlsbad, Calif., and the University of California, Los Angeles [J. Am. Chem. Soc., 117, 10787 (1995)]. This approach differs from the Scripps technique in that reaction data from each stage of combinatorial synthesis are stored in a computer database, rather than being retained in the chip itself. An identification number stored in the memory of each chip is a pointer to reaction information in the database. Moran and coworkers have applied the strategy successfully to the discovery of novel inhibitors of a protein tyrosine phosphatase.
A combinatorial chemistry application that has become increasingly active in the past year or so, and that promises to grow even more rapidly in the future, is combinatorial molecular recognition - the use of combinatorial techniques to study binding between biological or synthetic receptors and their ligands. Researchers in the combinatorial molecular recognition community "want to be able to make small molecules that do the kinds of things that antibodies do - tightly and selectively bind important molecules or transition states," explains Columbia's Still.
"We made libraries of substrates just to measure the binding properties of compounds synthesized as enantioselective receptors," says Still, "and the receptors did indeed have significant sequence-selective binding properties that had never been observed before." The results of such experiments suggest, says Still, "that virtually anything people can do with antibodies ought to be doable with small molecules, and that it may not be that hard to identify small molecules that are as selective as antibodies for binding substrates."
Chemistry professor Stuart L. Schreiber and coworkers at Harvard University are also using combinatorial molecular recognition - in this case in conjunction with nuclear magnetic resonance spectroscopy (NMR) - to study protein receptors. They have focused initially on the SH3 domain, a frequently occurring structural feature in proteins (such as tyrosine kinases) involved in signal transduction.
The researchers identified peptide ligands that bound SH3 in two binding pockets that make up part of the SH3 binding site. The SH3 binding site also includes a third binding pocket that is highly variable in structure and is therefore referred to as a "specificity pocket." A combinatorial strategy led to the discovery of two classes of peptide ligands that bind to the three pockets in opposite orientations, as determined by NMR analysis of the SH3-ligand complexes. Last month, Schreiber and coworkers reported also having identified nonpeptide elements that bind to the specificity pocket [J. Am. Chem. Soc., 118, 287 (1996)].
Chemistry professor Fredric M. Menger and coworkers at Emory University, Atlanta, are also using a form of combinatorial molecular recognition - in this case to identify industrial catalysts [J. Org. Chem., 60, 6666 (1995)].
"Libraries have in the past been screened for noncovalent binding," says Menger. "But this is the first, or one of the first, cases where purely organic libraries have been investigated for catalytic activity. We make hundreds or thousands of compounds very quickly and then test their catalytic power. The potential catalysts are polymers that have multiple functional groups in different proportions and different sequences, plus a metal ion."
In screening for catalytic activity, "we selected the hydrolysis of a phosphate ester," says Menger, "but one could choose any reaction of interest. Once a polymer with activity is found, we begin tinkering with the proportions to fine-tune it until it gets faster and faster."
Using this approach, Menger and coworkers have identified polymers that accelerate phosphate hydrolysis by a factor of 104 or more. According to Menger, "The potential exists for even greater acceleration ... since only a small portion of the vast number of possible combinations has as yet been tested."
In future work, the researchers hope to make chiral polymers that can reduce functional groups enantioselectively. "I would not be surprised if in 10 years most new catalysts are developed combinatorially," says Menger. "Industry is moving more and more toward aqueous systems to avoid organic solvents. If one could devise catalysts for organic reactions in water, that would be a useful practical development."
Planning and performing combinatorial experiments in the laboratory is a complex and potentially tedious process. Hence, "A future trend is going to be greater availability of automation devices," says Chiron's Zuckermann. "A lot of solutions are being developed for automating combinatorial split synthesis or multiple parallel synthesis."
For example, Chiron has developed proprietary robotic combinatorial synthesizers. "We now have third-generation units working in our labs that feature all-glass reaction vessels, heating to 120°C, and flexible software, [allowing] automation of most organic reactions," says Zuckermann.
Ontogen has developed OntoBLOCK, an in-house combinatorial chemistry automation system that can produce 1,000 to 2,000 small organic molecules per day by parallel array synthesis. The system includes reaction blocks containing 96 reaction vessels, from which compounds can be transferred directly to standard 96-well microtiter plates for high-throughput screening.
Bohdan Automation Inc., Mundelein, Ill., markets a combinatorial chemistry reaction block that accommodates a wide variety of organic solvents and handles both solid-phase and solution-phase chemistry. Advanced ChemTech, Louisville, markets instrumentation for combinatorial peptide and organic synthesis. And Tecan U.S. Inc., Research Triangle Park, N.C., offers an organic chemical synthesizer called CombiTec that includes a robotic sample processor and reaction blocks of eight to 56 chambers.
Robotics maker Zymark Corp., Hopkinton, Mass., has put together several different automation systems that enable their clients to do solution-phase combinatorial synthesis and solid-phase peptide and peptoid synthesis. The reactions can generally be performed under inert gas at a variety of temperatures.
According to Brian Lightbody, general manager of drug discovery business development at Zymark: "The process of generating combinatorial compounds involves several steps in addition to the actual reaction - [including] initial formulation of the reactants, labeling, pooling and splitting, cleavage, liquid-liquid extraction, solid-phase extraction, and evaporation. These steps require extensive manual labor. ... An automated robotic approach can often be implemented to fulfill these requirements, dramatically reducing the manual labor and eliminating the sources of human error."
A combinatorial chemistry system still in the prototype stage is the Nautilus, a synthetic chemistry workstation being developed by Argonaut Technologies Inc., San Carlos, Calif. The instrument handles a wide range of reagents, with capabilities for temperature control and use of inert atmospheres.
"The Nautilus allows you to do pretty much what you're able to do on the bench except in an automated fashion," says Argonaut President and CEO Joel F. Martin. "The system is completely enclosed and encapsulated, with a pressurized fluid delivery system and no exposure to the atmosphere whatsoever. It's a closed system, and all wetted surfaces within the instrument are glass or [polytetrafluoroethylene, such as DuPont's] Teflon."
Procedures that have been demonstrated on the Nautilus include a Suzuki coupling (a carbon-carbon bond-forming reaction at elevated temperature using an air-sensitive palladium catalyst), a butyllithium reaction, enolate reactions of the type developed by Ellman and coworkers, and synthesis of a solid-phase druglike molecule. "We chose tough organic reactions that no one would ever have conceived of doing in an automated synthesizer in the past," says Martin. The Nautilus is scheduled to be released commercially in August.
CombiChem Inc., San Diego, is developing commercial instrumentation for combinatorial chemistry that is likely to be competitive with the Nautilus. "Every company now is looking at ways of automating synthesis, purification, and analysis," says CombiChem Chief Operating Officer Peter L. Myers. "There's a major revolution going on. It's probably not obvious to a lot of people, and the academics may think we're overemphasizing it. But I know for a fact that every company now is looking to automate combinatorial chemistry because chemistry's become the rate-determining step."
As to whether conventional robotics instrumentation can be used effectively for combinatorial chemistry synthesis, "This immediately gets one into a debate," replies Myers. "When you're doing chemical reactions to make small molecules, many of the reactions are sensitive to conditions such as the presence of water vapor or oxygen, so inert atmospheres such as argon and nitrogen are often needed. The only successful way of blanketing a reaction is to have a closed system - one that is sealed. And if you seal it, then of course you can't use a robot very easily."
Myers adds, "This is why we, and also Argonaut, have gone to nonrobot systems - closed systems that work on valves and plumbing. ... That essentially means individual reaction vessels presealed with a solid support or chemicals inside, delivery by valving, and some way to agitate or stir the contents. Then you let the reaction proceed and wash the resin at the end, if it's a solid-phase reaction." However, he concedes that many researchers are currently using robotic systems instead of closed systems for combinatorial synthesis, "so the jury is out on which is the most acceptable."
CombiChem's instrument will be capable of automating both solid-phase and solution chemistry. "If you really want to exploit as much diversity as you can ... you have to be able to do something in addition to just solid-phase chemistry," says Myers. "The reason is pretty obvious. There are about 150 reactions now that work on solid phase. Some of those reactions work extremely well, some are still very poor yielding. But the organic chemists have an armory of thousands of chemical reactions that have been developed over the years, and of course primarily most of those are done in solution."
3-Dimensional Pharmaceuticals Inc., Exton, Pa., is also developing combinatorial chemistry instrumentation. The system is based on a technique called DirectedDiversity, an iterative optimization process that explores combinatorial space through successive rounds of selection, combinatorial synthesis, and testing. In each step, a chemical library is generated by robotic instruments, structure-activity information is obtained on library members, and data are analyzed to determine how closely the synthesized compounds match a set of desired properties. In each succeeding iteration, the structure-activity models are refined and new compounds are created until desired drug leads have been identified.
Combinatorial chemistry has come a long way in the past few years, but many challenges still lie ahead. For example, Ellman foresees further development of solid-support chemistry, including new linkage strategies and novel methods for synthesizing support-bound libraries and cleaving compounds from supports. "And people will continue to focus on different types of templates - novel templates for the versatile display of functionality," he says.
Ellman also believes "there are some interesting opportunities in the area of combining combinatorial strategies with computational strategies and structure-based design. The idea is to use information about three-dimensional structures of receptors and enzymes in combination with libraries to rapidly identify high-affinity ligands. It is going to be interesting to see how best to combine these two approaches." The recent SH3 study by Schreiber's group exemplifies this strategy of using a knowledge of protein and protein-ligand structure to help design optimal libraries.
Eric M. Gordon, vice president of research and director of chemistry at Affymax, points out that "some people enter into the molecular diversity sphere with the idea that it's a random process and that what you want to do is make as many molecules as you can that are as different from each other as they can be, but that no particular thought has to go into the design of these libraries. That's an extreme position. I believe that combinatorial chemistry doesn't stand alone. It should be integrated in with the arsenal of tools used for drug discovery, as opposed to being viewed as a competitive technology, say with structure-based design. I think the fate of it and the greatest power of it is going to be when it's used in concert with structure-based approaches and computational approaches."
Still believes that future prospects for combinatorial chemistry are good. "A lot of drugs will be discovered with it," he says. "And it's hard to imagine that there will not be people who find some enormously interesting catalytic compounds and stoichiometric reagents using these methods." However, he says, "The real key to making it work is twofold. First, you really need to have a good idea - a good basic structure that has a real chance of doing something really interesting, and you want to manufacture that idea in as many variants as you can afford to screen."
A second and even greater challenge, says Still, is devising novel and effective assays. "You need assays for the property you want that can be run in parallel, so you can select the beads or the library members that you want just by simple inspection. You simply look at them and pull out the ones that have the right property."
Gordon agrees that greater assay development is needed. "The amount of molecular diversity and the number of molecules that are going to become available are going to dwarf the present screening capacities," he says. "What's required is more individualized assays and assay miniaturization."
Miniaturization not only saves on the cost of proteins and other rare materials used in assays but also provides greater compatibility with the very small amounts of combinatorial compounds that are typically synthesized on beads. "Instead of 96-well microtiter dishes, which are the current standard in the pharmaceutical industry, you're going to see 1,000-well trays," Gordon predicts.
Ontogen's Moran believes that an increased focus on analytical chemistry is needed in the combinatorial field. "One needs to produce a reasonable amount of material in order to characterize any one compound that might be of interest in a library, either by mass spectroscopy or more preferably by proton NMR," says Moran. "The reason for this is that organic synthesis is not straightforward."
Different groups added to a core structure will affect the reactivity of library members to a differential extent, leading to possible failures of key synthetic steps. "So one needs to get a handle on how much of each component is being produced and whether you're actually making all the components in your library," he says. "Unless we have good analytical control over our experiments, it's going to be a challenge knowing what one's made." However, another researcher comments that, although it's important to be able to analyze hits, it's not practical or even desirable to analyze all library compounds.
Janda believes another key goal for the future is "to create a global library or universal library, where if you screen the library you'd find a hit for any type of target. It might not be a very potent hit, but you'd find a lead. People are trying to create a small library that would be very diverse that would give you leads to almost anything in the drug area. Some people call this combinatorial chemistry's Holy Grail."
However, Still says the universal library may prove to be as permanently elusive as the Holy Grail. In combinatorial chemistry, he says, "medicinal chemists ... design the first library of 1,000 to 100,000 compounds that has a good chance of acting on the target they are going after. Then they screen and make a new sublibrary based on the structure-activity relationships they find. I don't think any medicinal chemist would believe any single library of 10,000 compounds, no matter how carefully chosen, will contain leads for every medical target."
Schreiber suggests that combinatorial molecular recognition could become a fundamental tool for understanding protein function. One of the ultimate goals of the Human Genome Project is to discover the functions of human proteins, he says, and up to now this has been done with molecular biology techniques.
"Virtually all studies of the functions of proteins today involve making mutations in the genes that encode proteins and studying the effects," he explains. "This genetic approach to studying protein function is very powerful, but it is very slow and very inefficient. It's going to take centuries to study the function of all the proteins encoded by the human genome this way, and that's simply unacceptable."
In principle, this problem could be solved, he says, by using a "chemical genetics approach - where instead of making mutations in the gene encoding the protein you attack the protein itself by using organic ligands that bind to it." And such ligands can best be identified with combinatorial methods.
Hence, says Schreiber, "Chemical genetics could be the way in the future to solve the problem of protein function. There's a big advantage if you do it that way - because the very act of understanding protein function gives you a molecule that actually alters function. In terms of medical applications of the knowledge we seek, that's what one is ultimately trying to do."
Combinatorial chemistry, coupled to structural biology and cell biology, "is the most likely avenue to solve the protein binding problem," he says. "If we can combine those techniques, the consequences will be very exciting. It will lead to an era where biology is intimately coupled to chemistry, and where one might even say that chemistry, rather than genetics, will drive biology."
The ultimate usefulness of combinatorial chemistry for drug discovery and other applications remains to be proved. But Lilly's Kaldor - whose group developed by combinatorial means the CNS agent that has advanced to the clinic - is one researcher who is cautiously optimistic. "These techniques are more broadly applicable than crystal-structure-guided design methods because you don't have to have any knowledge of your receptor in order to apply them. ... You can develop a pharmacophore hypothesis much more quickly than you might have otherwise been able to do so. To date, we have used combinatorial chemistry for lead generation or lead optimization in over 50% of current Lilly projects and anticipate this percentage will increase with time."
Lilly's development of the CNS compound took less than two years from target identification to the beginning of clinical trials. This is "very fast," says Kaldor, "and we, of course, are being challenged by our management to repeat this success in every project we work on. ... It's a stunning example of what can be done if ... you apply combinatorial chemistry."
[ACS Home Page]
[ACS Publications Division Page]