C&EN Washington

Combinatorial chemistry is the art and science of synthesizing and testing compounds for bioactivity en masse, instead of one by one, the aim being to discover drugs and materials more quickly and inexpensively than was formerly possible. It has been the hottest approach to drug discovery long enough now that researchers have recognized some of its limitations and are developing strategies to address them.

"Most people would agree that it's difficult to get high-throughput chemistry to deliver drugs consistently," says Jonathan Knowles, head of global pharmaceutical research at Hoffmann-La Roche, based in Basel, Switzerland. "Sometimes it works, sometimes it doesn't. If it's difficult, why is it difficult?" he asks rhetorically.

"I think we've had the wrong goals. One of the attractions of parallel-synthesis chemistry, combinatorial chemistry, or whatever you want to call it, was numbers. Many of us believed that 'numbers' were the answer to diversity. Unfortunately, the diversity we created was often irrelevant, or only semirelevant, to the biology we were interested in interfacing with."

If each compound in a 10²⁴-compound library were loaded into the tiny wells of microtiter plates, Knowles notes, the resulting library would cover the U.S. to 30 times the height of the Empire State Building. "Assuming we could generate that much chemical diversity, it wouldn't really help because there would be no space for biologists to do the screens," he jokes. Hence, chemical diversity per se is not the answer, he says. "It's biologically relevant chemical diversity. The chemistry must be focused."

Knowles expressed these views at the 3rd Lake Tahoe Symposium on Molecular Diversity, held in late January in Lake Tahoe, Calif. At the symposium, scientists from an unusually wide range of disciplines--including chemistry, pharmaceutical research, molecular biology, and medicine--discussed some of the latest techniques and strategies in solid-phase synthesis, library design, informatics, screening, genetic analysis, and other areas.

According to Knowles, selecting an appropriate target--the human protein at which a drug's activity is directed--is more important than sheer numbers in combinatorial drug discovery. If you get the target wrong, he says, "nothing you can do afterward will give you a medicine. That's why 90% of what happens in research and development--at least in any of the companies I've ever worked in--is a complete and utter waste of time. That's not quite true--it's fun. But it doesn't lead to medicines."

Also required for successful combinatorial drug discovery, he says, "are innovative reactions to generate the types of molecules we want in parallel synthesis, which probably means spending more time making building blocks. I believe that we need better in-silico [computer-aided] design--not just to design drugs but to design libraries, because I can't see any other way out of the Empire State Building problem. We need accurate algorithms to predict protein structures and molecular interactions. We need extensive data stored in a common format that describes all the aspects of a molecule." Informatics--the use of computers to analyze complex data--is therefore "changing and will change the way we do science," he says.

"We need ultra-high-throughput screens with multidimensional readout systems to define the relevant diversity. It isn't enough to be doing our in-silico stuff--we'd better be able to test it in real time, and we'd better be able to test it with surrogate biological systems. And we need talented, flexible people who are capable of talking to each other about their disciplines in order to use this knowledge to continuously improve their actions."

Knowledge and understanding, Knowles adds, "will be generated, I believe, through an understanding of diversity. Understanding biological diversity gives us better targets, understanding chemical diversity gives us better drugs, and understanding human diversity gives us better health care."

C. Fred Fox, a professor in the department of microbiology and molecular genetics at the University of California, Los Angeles, and a member of the group that organized the Lake Tahoe symposium, agrees with Knowles on the need for collaborative efforts. "It's important to look at the whole picture--the chemistry, the screen, the biology that drives the screen, and the informatics that holds these things together," Fox says. "The parts aren't worth very much without the whole. Some of the new high-throughput technologies are very powerful, but unless they're applied together and cross-validated, you can't take full advantage of what each can offer and you won't achieve all that can be realized."

Combinatorial chemistry has become increasingly popular in academia, government, and industry in the past few years, but it's still a relatively young set of technologies and its value remains unproven. "Initially, some skeptics said, 'You can make and screen a lot of compounds, but I haven't seen the fruit of the labor yet,' " says Suvit Thaisrivongs, director of chemistry at Pharmacia & Upjohn, Kalamazoo, Mich. "Now we are starting to see some, but are these isolated incidents or is this going to be the norm? Everyone would like to reduce the time it takes to discover a drug. But can we really expect the delivery of drugs to be faster and more consistent as we all invest in new technologies? That's a question all of us would like to know the answer to."

Combinatorial chemistry does already appear to be paying benefits in some cases, as demonstrated by two seemingly very different routes to drug discovery report-ed at the symposium by researchers at Merck, Rahway, N.J., and CombiChem, San Diego. The independent studies show "that combinatorial chemistry has come of age," Fox says. "They demonstrate how combinatorial strategies can be applied rationally, deliberately, and successfully."

Thaisrivongs concurs, saying the Merck and CombiChem studies represent "the current state of the art of combinatorial chemistry--how you can use it to go from hits to potential clinical candidates."

At Merck, research fellow Scott C. Berk, senior research fellow Susan P. Rohrer, and coworkers used a combinatorial approach to identify nonpeptide agonists (activators) for each of five different somatostatin receptor subtypes. Their study "never could have been done so quickly without combinatorial chemistry," says chemistry professor Ole Hindsgaul of the University of Alberta, Edmonton. Using conventional techniques, "it would have taken 10 years," he surmises.

Somatostatin is an endogenous peptide that plays important physiological roles as an inhibitor of hormone secretion in the pituitary, pancreas, and gastrointestinal tract; as a neurotransmitter in the brain; and as an inhibitor of tumor cell growth. Researchers therefore consider the five distinct human somatostatin receptor subtypes (sst1 through sst5) to be promising targets for drug discovery.

"Somatostatin doesn't distinguish among the five different receptor subtypes--it binds with high affinity to all five," Rohrer says. "So when you observe a physiological response that's due to somatostatin, you really don't know which receptor subtype has mediated that effect." Hence the functions of the different receptor subtypes are understood incompletely at best.

The Merck group was interested in finding somatostatin analogs that could bind selectively to the different receptor subtypes--not only to identify potential drug leads, but also to help them better understand the specific physiological functions of each receptor subtype. Although analogs selective for sst2 had been found, compounds that bound to the other subtypes had not.

By combinatorial means, Berk, Rohrer, and coworkers were able to identify nonpeptide agonists for each of the five subtypes [Science, 282, 737 (1998)]. The researchers first searched the Merck chemical database for small molecules similar in structure to a somatostatin agonist. The compound that turned out to bind with highest affinity to somatostatin was then used as a template for library construction. The Merck team synthesized the libraries on solid support beads by split-and-pool synthesis, a multistage process for producing very large numbers of compounds. They then cleaved the compounds from the beads and screened them as mixtures in solution. Working in solution, Berk says, "allowed us to use standard assay technology, identical to assays used to screen single compounds, instead of having to develop a custom on-bead screen."

The researchers synthesized the libraries as mixtures of large numbers of compounds, instead of as individual discrete compounds. This mixture approach has been generally out of favor among combinatorial chemists in the past few years because it's considered very difficult to figure out what component or components are contributing to the activity detected in a complex mixture.

The Merck group solved this problem by using an established technique called deconvolution. Compounds responsible for receptor subtype-selective activity in the screened mixtures were identified by backtracking through the resin pools that had been used to synthesize the mixtures. Berk, Rohrer, and coworkers had saved those pools after each library synthesis step.

After the subtype-selective compounds had been identified, the researchers used them to define physiological roles for some of the receptor subtypes [Proc. Natl. Acad. Sci. USA,95, 10836 (1998)]. For example, they found that the sst2 and sst5 receptor subtypes regulate growth hormone release in rat pituitary glands.

The Merck combinatorial approach is "one of the better examples to date where mixture screening has produced viable candidates with the desired selectivity profile going in," says research scientist David Mendel of Eli Lilly, Indianapolis. Mendel believes mixture-screening technology has improved recently and that the technique is currently "making a comeback."

A contrasting approach has been adopted by Chief Scientific Officer Peter L. Myers and coworkers at CombiChem, who are finding biologically active agents in small, highly directed combinatorial libraries instead of in complex mixtures. At CombiChem, Myers says, "we're very focused in terms of the numbers of compounds we make. We use parallel synthesis coupled with drug design technology to reengineer the drug discovery process." The researchers try to identify important property differences between active and inactive molecules and then enrich their libraries in compounds having the favorable characteristics.

"That was the basis of rational drug design in the '80s," Myers says, but "the problem with that approach was that it had to be a hole in one. You took the combination of features that looked perhaps the most relevant and then you tried to express them all in one molecule. Then you'd go make that molecule, and if it was a lemon, you kind of were no wiser. We use libraries, so we're automatically looking at larger numbers of compounds and we're not expecting to get an exact match. We're just expecting to see some marginal improvement in activity of some of those compounds" as the focused libraries are synthesized, screened, and analyzed in an iterative manner.

With the focused library concept, Myers says, it takes six to eight cycles at two to three months per cycle to go from hits to leads--compounds suitable for development as drugs. The average time for lead generation is thus 18 months--unusually fast by conventional drug development standards. Myers notes that CombiChem has used this strategy to discover promising compounds that interact with a member of the chemokine receptor family and to identify inhibitors of a specific metalloproteinase enzyme. The lead compounds coming out of these studies are of potential utility in the treatment of chronic inflammatory diseases, such as rheumatoid arthritis.

"In contrast to the Merck example, where large libraries of compound mixtures were synthesized and screened," Myers says, "the CombiChem process requires modest numbers of compounds in synthesis and screening--less than 5,000 in total--to achieve drug development candidates with the desired clinical profile." The CombiChem studies show, he says, that "small, focused libraries of single compounds can be very effective in lead discovery against a range of new biological targets."

He believes directed libraries are preferable to what he calls "the shotgun approach"--the idea that "if you can make and screen millions of compounds you'll find a lot more drugs." Research groups at large corporations "may be able to make that work," he says, "but it's an expensive process."

Nevertheless, UCLA's Fox comments that the Merck and CombiChem approaches are each appropriate to the companies' different circumstances, aside from economics. In Merck's case, he says, "a vast history of multidisciplinary studies on the targets and leads was already at hand, [whereas] the CombiChem approach is more typical of what is necessary when you're presented with a target that has little to no history behind it."

Fox notes that "Merck has tens of thousands to hundreds of thousands of potential lead compounds derived from its 100 or so years of medicinal chemistry experience, and these are cellared away like good wine. CombiChem"--and other young companies such as Axys Pharmaceuticals (South San Francisco), Signal Pharmaceuticals (San Diego), Pharmacopeia (Princeton, N.J.), and Alanex (San Diego)--"is starting from scratch and has to create a library of potential leads without the benefit of that history. Once you get past this major difference," both large and small companies "are likely to be speeding along the same freeway," he says.

New strategies for synthesizing libraries on solid-phase beads are also coming down the pike. "There are innumerable synthetic methods, but not all of them are applicable to solid-phase reaction conditions," says Shripad Bhagwat, director of medicinal chemistry at Signal Pharmaceuticals. "This is a very critical area of interest in combinatorial chemistry because solid-phase reactions make it possible to prepare libraries of more diverse structure."

Robert W. Brueggemeier, a professor of medicinal chemistry at Ohio State University, Columbus, adds: "A lot of people are now adapting solution-phase chemistry to solid phase and expanding the horizon of organic chemistry that can be done on solid support."

One of those people is professor of pharmaceutical sciences Sh Kobayashi of the University of Tokyo. Kobayashi is developing methods for library synthesis in which catalysts--rather than reagents and products, as usual--are immobilized on solid supports. Polymer-supported catalysts can be very easily removed from solution after synthetic reactions have been completed. And supporting catalysts instead of reagents eliminates problems such as slow reaction rates and variations in reactivity that sometimes arise from reagent immobilization.

Kobayashi has used polymer-supported polyallylscandium trifylamide ditriflate to synthesize large numbers of quinoline derivatives. The syntheses were carried out simply by mixing the polymer-supported catalyst with an aldehyde, an aromatic amine, and an alkene or alkyne in an organic solvent.

"After the reaction," Kobayashi says, "the filtrates are concentrated to give almost pure quinoline derivatives in most cases. Hundred-milligram-scale syntheses of a large array of diverse molecular entities have been achieved with high purities, high yields, and high selectivities. With this method, a quinoline library of more than a million compounds could be prepared using an automated system."

The activity of polymer-supported catalysts is usually lower than that of corresponding monomeric catalysts. However, Kobayashi and coworkers have developed a polymer-supported Lewis acid--a microencapsulated scandium triflate--that is even more active than the corresponding monomeric scandium triflate catalyst.

The catalyst-immobilization technique developed by the group is a new method "based both on physical envelopment [of the catalyst] by the polymer and on electronic interactions between electrons of polymer benzene rings and a vacant orbital of the Lewis acid," Kobayashi explains. The catalyst is readily prepared, maintains high activity, is recoverable and reusable, and can be successfully applied to many useful carbon-carbon bond-forming reactions. He and his coworkers have also used the technique to prepare a microencapsulated osmium tetroxide catalyst.

Traditionally, scandium triflate and osmium tetroxide catalysts have been a nuisance to remove from solution after reactions. So Kobayashi's microencapsulated versions could be very helpful to synthetic chemists.

Meanwhile, chemistry professor Shankar Balasubramanian and coworkers at the University of Cambridge, in England, have been studying cyclization reactions for use in solid-phase synthesis, such as a stereoselective cyclization for preparation of oxopiperazines that exhibits a high degree of stereoselectivity. In this reaction, carried out by graduate student Nawaz Khan, a mixed diastereomeric intermediate bound to a solid support partitions in two directions. One diastereomer cyclizes on the solid-phase resin to generate a pure stereoisomeric product bound there, while another diastereomer kinetically prefers to undergo a cyclative cleavage that removes it from the resin. Hence, the reaction cleans itself up to produce just one pure stereoisomer on the resin, even when the precursor is a mixture of isomers--a result that can be very useful in library design.

Balasubramanian and coworkers are also developing new reactions for chemical tagging of split-and-pool libraries, such as the use of Friedel-Crafts chemistry to chemically encode solid-phase-bound libraries. "We're trying to achieve improved selectivity in tagging without the aid of any protecting group chemistry," Balasubramanian says. "So far we've exemplified it with some peptide libraries that we've encoded."

A major difference between combinatorial solid-phase synthesis and classical solid-phase synthesis of a single target molecule, Balasubramanian says, "is that in the combinatorial case it is insufficient to optimize the chemistry in just one context. We need to identify reactions that are going to work equally reliably in a multitude of chemical contexts, and there aren't many reactions that are known to do that at the moment. I think a way to address that will be to carry out systematic studies in a statistically significant number of cases for a given piece of chemistry carried out on solid phase."

Balasubramanian notes that a related approach was reported at the symposium by Peter Schneider of the metabolism and cardiovascular diseases unit of Novartis Pharma, Basel. Schneider has carried out a systematic study of the effects on solid-phase chemistry of size and property differences in supposedly homogeneous batches of resin beads used as solid-phase supports.

He found that the homogeneity of size, shape, and surface functional groups of beads can have significant effects on the reproducibility and yield of solid-phase reactions. According to Schneider, these properties vary considerably within and among batches of commercially available beads used commonly in combinatorial synthesis. His study found that the amount of compound on a single bead varied by up to a factor of 10.

Schneider has found only one company-- LCC Engineering & Trading, based in Egerkingen, Switzerland--that has demonstrated the ability to make solid-support beads that are much more homogeneous than typical commercial beads. The company uses a proprietary polymerization process to produce the highly uniform and consistently functionalized beads, which it calls LCC-Dynospheres. Schneider uses the beads for one-bead/one-compound synthesis and finds the chemistry to be more reproducible than with conventional beads.

"You don't normally see discussions or publications on the importance of bead size and how it really impacts the efficiency of the chemistry on solid supports," Ohio State's Brueggemeier comments. "Schneider took the initiative to really look at the effect of bead properties on reaction rates, reaction conditions, and yields, and he pointed out that the nonhomogeneous nature of beads really can have an impact on how efficient and how effective your chemistry is. Most of us just buy beads off the shelf and use what we get. Sometimes reactions go well and sometimes they don't, and you're not really sure why things aren't working that well."

Balasubramanian adds: "For me, the message was for us to try to be more scientific in the way we approach solid-phase organic synthesis. At the moment, like a great deal of organic synthesis, it is a skillful art form to some extent, and we need to reduce it to more of a science as far as we can. I think for too long we've been trying to see what's possible in solid-phase synthesis rather than getting at the root of why, for example, reactions aren't always efficient on solid phase. Schneider's talk was an example of trying to standardize the parameters [and reduce] the complexity of getting solid-phase chemistry to work across various compounds and contexts, which is clearly a struggle."

One way to design libraries is to make them "dynamic." Evolution provides the inspiration for dynamic combinatorial chemistry, says assistant professor of medicinal chemistry Alexey V. Eliseev of the State University of New York, Buffalo. In dynamic libraries, components exist in dynamic equilibrium and can easily interconvert with each other. When an immobilized target compound is introduced, some library compounds bind to it selectively and are removed from solution. The equilibrium then shifts, amplifying the good binders and minimizing the concentration of poor binders in the library--a form of test-tube evolution. Several groups are currently studying dynamic combinatorial chemistry [Angew. Chem. Int. Ed.,37, 2828 (1998)].

Eliseev believed he had demonstrated the principle for the first time in 1997 when he showed it was possible to isolate and enrich one compound in a three-member library that bound with highest affinity to a target. The compound was one of three dicarboxylates that could interconvert with each other.

After the dicarboxylate with the best binding activity was removed from the mixture by affinity chromatography, it was regenerated by isomerization of the other two compounds. Repeated removal and regeneration of the best binder thus made it possible to enrich and identify the compound. Eliseev later found out, and now emphasizes, that medicinal chemistry professor Duane L. Venton and coworkers at the University of Illinois, Chicago, reported a similar approach (but in an entirely different system) one year earlier, in 1996.

Eliseev's initial work on dynamic combinatorial chemistry was done in organic solvents, where it was easier to carry out, "but now we're moving to biologically relevant systems," he says. "We've developed chemistry for dynamic libraries that can work in aqueous solutions. That's essential because selecting for biological targets is done in water." His group currently has a paper in press in the Journal of Physical Organic Chemistry on the use of imine exchange reactions in dynamic libraries of oxime ethers in aqueous solution.

"Amplifying components with affinity selection from a pool of compounds in dynamic equilibrium is a very good idea," Pharmacia & Upjohn's Thaisrivongs comments. "It's still a long way from being practical and successful, but the concept is there. Not that many chemical systems are amenable to this"--because all the components of a dynamic library have to be in equilibrium with each other. "So it is probably useful only for a very small subset of possible reactions. But selection by chemical evolution could be a very powerful tool down the road."

An important advantage of dynamic libraries, Eliseev says, is that "they simplify mixture screening, and mixtures can be created much more easily than libraries of individual compounds. What people mostly do in companies now is parallel synthesis--one compound per bead or one compound per well--and parallel screening. That's good because you know what you're dealing with, but it takes a lot of time. You can spend months or years on parallel library synthesis. You can make mixtures in hours or days," but elaborate techniques are needed to identify active components in complex mixtures, he says. "Our approach eventually intends to simplify working with mixtures by changing their composition so they have higher percentages of hits, or better binders."

Eliseev's group also is trying to develop a complementary approach for dealing with mixtures--a special way to tag mixtures for identification purposes by labeling molecular scaffolds (compounds used to create libraries) with different fractions of ¹⁵N. The project "hasn't been completed yet," Eliseev says, "but we probably will publish on the technique later this year."

If one accepts Knowles's notion that numbers aren't everything, then it's important to have rational, computational techniques for designing the diversity of libraries and analyzing the three-dimensional shapes of synthetic products. One example of such technology is a new software algorithm that can maximize diversity and minimize cost in library design. The program was developed by research scientist Karl M. Smith and Robert S. Pearlman, director of the Laboratory for Molecular Graphics & Theoretical Modeling at the College of Pharmacy of the University of Texas, Austin.

Consider a simple combinatorial library synthesis--the reaction of a set of type-A reactants with a series of type-B reactants. One might react the 536 commercially available secondary amines found in a chemicals database with the 2,061 purchasable aldehydes listed there to form a "virtual" library with a total population that could potentially exceed a million compounds.

To save the time and expense required to synthesize and screen more than a million combinatorial products, pharmaceutical company researchers would typically choose to synthesize a small, diverse, representative subset of the virtual library--say, just 9,600 compounds. The question is, which 9,600 out of the million or so should they actually make?

To make the 9,600-member library, Pearlman says, "the traditional approach is to select, for example, the 80 most diverse of the 536 available secondary amines and the 120 most diverse of the 2,061 available aldehydes. While this approach is very economical, requiring relatively few reactants, the diversity of the resulting 9,600 compounds is quite low compared with the maximally diverse library one could select."

The new software developed in Pearlman's lab makes it possible to select an optimal set of reagents to maximize library diversity and yet minimize reagent cost. The software, based on a "reactant-biased, product-based library design" algorithm, will be available commercially as a feature of DiverseSolutions, a combinatorial chemistry and diversity software suite developed by Pearlman and coworkers and marketed by Tripos, St. Louis. Pearlman notes that the algorithm works equally well for designing diverse libraries and focused libraries.

New methods are also being developed in the area of high-throughput library screening. Illumina, San Diego, is developing self-assembled bead arrays on fiber-optic-bundle tips that can detect multiple compounds in solution. Chemistry professor David R. Walt of Tufts University, Medford, Mass., discovered that fiber-optic bundles can be etched and then filled with beads.

Anthony W. Czarnik, chief scientific officer at Illumina, explains that fiber-optic bundles can contain thousands of individual fiber strands within a diameter of only 1 mm. A chemical treatment is used to etch tiny femtoliter wells in the end of each fiber. The wells are then filled with derivatized beads. The arrays are quite versatile in that the beads can be derivatized with a variety of reagent types, such as antibodies for immunoassay applications, DNA for genomic analysis, or fluorescent reagents for detecting reaction products in screening procedures. Illumina is initially developing self-assembled arrays for single-nucleotide polymorphism genetic analysis.

Insofar as potential high-throughput screening applications are concerned, Thaisrivongs says, "the Illumina technology is visionary in that it's trying to push the envelope further and further out. I'm quite interested in seeing how much the company can push the miniaturization to increase the number of assays that can be handled cost-effectively in a small space."

Fox notes that "every component in the arrays can be addressed individually and continuously in real time, which has lots of advantages for screening and validation. I think this is going to be a grand new addition to the armamentarium."

A method for high-throughput screening of natural-product-like libraries in a nanowell format has been developed by Randall W. King, institute fellow at the Institute of Chemistry & Cell Biology at Harvard Medical School. King uses extracts from Xenopus frog eggs to screen for small molecules that inhibit mitosis. A Xenopus extract is "a biochemical system that allows you to recapitulate many different types of cell biological events in vitro," King says. "This system has been used to model regulation of the cell cycle for some time, and what we've done is to adapt that to a high-throughput screening format."

Inhibitory compounds are identified with the aid of a reporter protein--a fusion product of part of the protein cyclin B with the chemiluminescence agent luciferase. In a normal cell cycle, cyclin B is degraded when cells exit mitosis. So in a screening experiment, if the extracts undergo mitotic biochemical changes, the reporter protein is degraded and there is a low signal. But if an inhibitor blocks any of the biochemical steps that precede or accompany mitosis, the reporter protein isn't degrad-ed and a bright luciferase signal is observed.

"Our assay looks for small molecules that block cell division," King says. "Inhibition can occur at any of several different transition points during the regulation of cell division. We start with a very broad assay, and once we find an inhibitor we then have to go back and figure out exactly what component is being inhibited."

King has miniaturized the assay for use in 1,536-well plates (2-L wells) and 6,144-well plates (250-nL wells). "This is getting down to a level that's compatible with screening one-compound-per-bead, million-member libraries," he says. "We're just now getting to the point where all this technology has been put in place, and we're starting to screen libraries."

The libraries that will be screened are large collections of natural-product-like compounds designed and synthesized by institute fellow Michael A. Foley and coworkers at Harvard. In the syntheses, Foley's group "used a set of very reliable high-efficiency reactions," Balasubramanian comments, including a tandem acylation and nitrone cycloaddition, butyrolactone aminolysis, esterification, and epoxide opening. "They approached this in a very academic way, and showed what was possible using a carefully designed scaffold and simple, elegant chemistry."

Fox points out that the Xenopus-extract-based screening technique is relatively inexpensive and thus practical for use both in academia and in the pharmaceutical industry. The Harvard work "suggests that academics can at this point begin to consider how they can truly become engaged" in high-throughput screening, an area that has not attracted great academic interest up to now, Fox says.

Of course, once active compounds are identified by high-throughput screening, a significant bottleneck to drug development still remains: How does one identify the compounds with the best chances for success in clinical trials?

George Grass, president of NaviCyte, a wholly owned subsidiary of Trega Biosciences, San Diego, proposes a unique solution--the use of in vitro permeability studies and computer modeling to predict the absorption and metabolic properties of compounds in people. Such properties have typically been estimated using animal models, which are expensive, slow, and generally unsuitable for evaluating large numbers of compounds.

NaviCyte researchers grow cells on membranes and then measure absorption by seeing how quickly compounds penetrate through the cells. Information from such experiments is then combined with computer simulation data to predict the compounds' success as drugs. Grass says he hopes the company will ultimately be able to base such predictions solely on chemical structure information, bypassing the in vitro testing steps entirely and thus speeding things along even further.

A number of Lake Tahoe presentations alluded to the growing connection between genomics and combinatorial chemistry. One crucial part of this relationship is that genomics can help researchers identify new molecular targets that they can then try to hit with their combinatorial tools. But genomics is likely to influence combinatorial chemistry in other ways as well over the coming years.

Microelectronic arrays developed by Chief Technical Officer Michael J. Heller and coworkers at Nanogen, San Diego, can be used for functional genomics assessments--tests to determine genetic differences that affect individual predispositions to disease and variations in the efficacy and toxicity of drugs.

And a group led by Peter Wilding, director of clinical chemistry at the University of Pennsylvania Medical Center, is developing microfluidics chips that can be used to isolate white blood cells from blood and analyze them to detect genetic sequences associated with inherited diseases.

So what's the connection to combinatorial chemistry? "There would seem to be very little," Wilding says. "When I started my lecture I expressed a little bit of concern that I was outside my field. However, the more I listened to the topics at that meeting, the more I became aware that microfabricated devices perhaps will eventually provide ways to screen the action of candidate compounds--to see whether they inhibit or enhance DNA replication and other cell processes. To do this, you would need to be able to distribute cells and reagents, carry out reactions, and detect reaction products, and all of this is becoming possible on microfabricated devices."

Wilding says that's why SmithKline Beecham, Brentford, England, has a collaboration with the microfluidics company Orchid Biocomputer, Princeton, N.J.; Hoffmann-La Roche, Nutley, N.J., and others have made deals with the lab-on-a-chip company Caliper Technologies, Mountain View, Calif.; and Becton Dickinson, Franklin Lakes, N.J., has an agreement with Nanogen. "The aim is to have the ability to do high-throughput screening in microfabricated devices," he notes.

Balasubramanian says the Lake Tahoe meeting "showed that the process of interfacing genetics and combinatorial chemistry is beginning. For me, there's still some way to go with bringing the two together in a clear and coherent way. But I guess there's been some good progress with that, and there will be a lot more to come in the future."

The Lake Tahoe symposium combined "outspoken people from different disciplines, and that to me represents the future of science and medicine," Knowles adds. "In the past, you could be seriously successful in science in your own little box. I think those times are over, and I think the future of science looks rather like this meeting."   

Chemical & Engineering News
Copyright © 1999 American Chemical Society