combinatorial chemistry and new drugs

Combinatorial Chemistryand New Drugs

An innovative technique that quickly produces large numbers of structurally related compounds

is changing the way drugs are discovered

by Matthew J. Plunkett and Jonathan A. Ellman

Copyright 1997 Scientific American, Inc.

To fight disease, the immune sys-tem generates proteins knownas antibodies that bind to in-

vading organisms. The body can makeabout a trillion different antibodies, pro-duced by shuffling and reshuffling theirconstituent parts. But the immune sys-tem is not equipped to craft a special-ized antibody each time it is faced witha new pathogen. Instead the body selec-tively deploys only those existing anti-bodies that will work most effectivelyagainst a particular foe. The immune sys-tem does this, in effect, by mass screen-ing of its antibody repertoire, identify-ing the ones that work best and makingmore of those. In the past few years, weand other chemists have begun to fol-low nature’s example in order to devel-op new drugs. In a process called com-binatorial chemistry, we generate a largenumber of related compounds and thenscreen the collection for the ones thatcould have medicinal value.

This approach differs from the mostcommon way pharmaceutical makersdiscover new drugs. They typically be-gin by looking for signs of a desired ac-tivity in almost anything they can find,such as diverse collections of syntheticcompounds or of chemicalsderived from bacteria, plantsor other natural sources.Once they identify a promis-ing substance (known in thefield as a lead compound),they laboriously make manyone-at-a-time modificationsto the structure, testing aftereach step to determine howthe changes affected the com-pound’s chemical and biolog-ical properties.

Often these proceduresyield a compound havingacceptable potency and safe-ty. For every new drug thatmakes it to market in thisway, however, researchersquite likely tinkered with andabandoned thousands ofother compounds en route.The entire procedure is time-consuming and expensive: ittakes many years and hun-dreds of millions of dollarsto move from a lead com-pound in the laboratory to a bottle of medicine on theshelf of your local pharmacy.

The classical approach hasbeen improved by screening

tests that work more rapidly and reli-ably than in the past and by burgeoningknowledge about how various modifi-cations are likely to change a molecule’sbiological activity. But as medical sciencehas advanced, demand for drugs that canintervene in disease processes has esca-lated. To find those drugs, researchersneed many more compounds to screenas well as a way to find lead compoundsthat require less modification.

Finding the Right Combination

Combinatorial chemistry responds tothat need. It enables drug research-

ers to generate quickly as many as sever-al million structurally related molecules.Moreover, these are not just any mole-cules, but ones that a chemist, knowingthe attributes of the starting materials,expects will have a desired property.Screening of the resulting pool of com-pounds reveals the most potent vari-eties. Combinatorial chemistry can thusoffer drug candidates that are ready forclinical testing faster and at a lower costthan ever before.

Chemists make combinatorial collec-tions, or libraries, of screenable com-

pounds in a rather simple way. We relyon standard chemical reactions to as-semble selected sets of building blocksinto a huge variety of larger structures.As a simplified example, consider fourmolecules: A1, A2, B1 and B2. The mol-ecules A1 and A2 are structurally relat-ed and are thus said to belong to thesame class of compounds; B1 and B2belong to a second class. Suppose thatthese two classes of compounds can re-act to form molecules, some variant ofwhich we suspect could produce a po-tent drug. The techniques of combina-torial chemistry allow us to constructeasily all the possible combinations: A1-B1, A1-B2, A2-B1 and A2-B2.

Of course, in the real world, scientiststypically work with much larger num-bers of molecules. For instance, we mightselect 30 structurally related compoundsthat all share, say, an amine group(–NH2). Next, we might choose a sec-ond set of 30 compounds that all con-tain a carboxylic acid (–CO2H). Then,under appropriate conditions, we wouldmix and match every amine with everycarboxylic acid to form new moleculescalled amides (–CONH–). The reactionof each of the 30 amines with each of

the 30 carboxylic acids givesa total of 30 × 30, or 900,different combinations. If wewere to add a third set of 30building blocks, the totalnumber of final structureswould be 27,000 (30 × 30 ×30). And if we used morethan 30 molecules in each set,the number of final combina-tions would rise rapidly.

Drugmakers have two ba-sic combinatorial techniquesat their disposal. The first,known as parallel synthesis,was invented in the mid-1980s by H. Mario Geysen,now at Glaxo Wellcome. Heinitially used parallel synthe-sis as a quick way to identifywhich small segment of anygiven large protein bound toan antibody. Geysen generat-ed a variety of short proteinfragments, or peptides, bycombining multiple aminoacids (the building blocks ofpeptides and proteins) in dif-ferent permutations. By per-forming dozens or sometimeshundreds of reactions at thesame time and then testing

Combinatorial Chemistry and New Drugs Scientific American April 1997 69

MIXING AND MATCHING of molecular building blocks inthe technique known as combinatorial chemistry allows research-ers to generate huge numbers of structures quickly. A robot(above) delivers the reactive chemicals used to assemble a largecollection of compounds (opposite page).C

HA

RLES

O’R

EAR

CH

ARL

ES O

’REA

R


to see whether the resulting peptideswould bind to the particular antibodyof interest, he rapidly found the activepeptides from a large universe of possi-ble molecules.

In a parallel synthesis, all the prod-ucts are assembled separately in theirown reaction vessels. To carry out theprocedure, chemists often use a so-calledmicrotitre plate—a sheet of molded plas-

tic that typically contains eight rows and12 columns of tiny wells, each of whichholds a few milliliters of the liquid inwhich the reactions will occur. The ar-ray of rows and columns enables work-ers to organize the building blocks theywant to combine and provides a readymeans to identify the compound in aparticular well.

For instance, if researchers wanted to

produce a series of amides by combin-ing eight different amines and 12 car-boxylic acids using the reactions we de-scribed earlier, they would place a solu-tion containing the first amine in thewells across the first row, the secondamine across the second row, and so on.They would then add each of the car-boxylic acids to the wells sequentially,supplying a different version to eachcolumn. From only 20 different build-ing blocks, investigators can obtain a li-brary of 96 different compounds.

Chemists often start a combinatorialsynthesis by attaching the first set ofbuilding blocks to inert, microscopicbeads made of polystyrene (often re-ferred to as solid support). After eachreaction, researchers wash away anyunreacted material, leaving behind onlythe desired products, which are still teth-ered to the beads. Although the chemi-cal reactions required to link compoundsto the beads and later to detach themintroduce complications to the synthe-sis process, the ease of purification canoutweigh these problems.

In many laboratories today, robotsassist with the routine work of parallelsynthesis, such as delivering smallamounts of reactive molecules into theappropriate wells. In this way, the pro-cess becomes more accurate and less te-dious. Scientists at Parke-Davis con-structed the first automated method forparallel synthesis—a robotic device thatcan generate 40 compounds at a time.And investigators at Ontogen have de-veloped a robot that can make up to1,000 compounds a day. In general, thetime needed to complete a parallel syn-thesis depends on how many com-pounds are being produced: when mak-ing thousands of compounds, doublingthe number of products requires nearlytwice as much time. Such practical con-siderations restrict parallel synthesis toproducing libraries containing tens ofthousands of compounds rather thanmany more.

Split and Mix

The second technique for generatinga combinatorial library, known as

a split-and-mix synthesis, was pioneeredin the late 1980s by Árpád Furka, nowat Advanced ChemTech in Louisville,Ky. In contrast to parallel synthesis, inwhich each compound remains in itsown container, a split-and-mix synthe-sis produces a mixture of related com-pounds in the same reaction vessel. This

Combinatorial Chemistry and New Drugs70 Scientific American April 1997

Parallel Synthesis

POLYSTYRENE BEAD

A1 MOLECULE

B1 MOLECULE

B2

B3

A2A3

STEP 1 Start with a molded plastic sheet of wells(called a microtitre plate) that have beenpartly filled with a solution containing in-ert polystyrene beads (gray circles). Typicalmicrotitre plates have eight rows and 12columns, for a total of 96 wells; this exam-ple illustrates only the upper left corner ofthe plate.

STEP 2 Add the first set of molecules—the A class (squares)—to the beads, puttingthe A1 molecules into the wells of the firstrow, the A2 molecules into the wells of thesecond row, and so on. After one set ofmolecules has been added, filter the mate-rial in every well to eliminate unreactedchemicals (those unattached to beads).

STEP 3 Add the second set of molecules—the Bclass (triangles)—down the columns, withB1 in the first column, B2 in the second, etcetera. Filter away unreacted chemicals asin step 2.

STEP 4 Once the 96-member library has been pro-duced, detach the final structures from thebeads so that the compounds can bescreened for biological activity.

JARE

D S

CH

NEI

DM

AN

DES

IGN


method substantially reduces the num-ber of containers required and raisesthe number of compounds that can bemade into the millions. The trade-off,however, is that keeping track of suchlarge numbers of compounds and thentesting them for biological activity canbecome quite complicated.

A simple example can explain this ap-proach. Imagine that researchers havethree sets of molecules (call them A, Band C), each set having three members(A1, A2, A3; B1, B2, B3; and so on). In-side one container, they attach the A1molecules to polystyrene beads; in asecond container A2 molecules, and ina third A3 molecules. Then the workersplace all the bead-bound A moleculesinto one reaction vessel, mix them welland split them again into three equalportions, so that each vial holds a mix-ture of the three compounds. The re-searchers then add B1 molecules to thefirst container, B2 to the second, and B3to the third. One more round of addi-tions to introduce the C molecules pro-duces a total of 27 different compounds.

To isolate the most potent of thesestructures, scientists first screen the finalmixtures of compounds and determinethe average activity of each batch. Then,using a variety of techniques, they candeduce which of the combinations inthe most reactive batch has the desiredbiological activity.

A number of pharmaceutical compa-nies have also automated the split-and-mix procedure. One of the earliest an-nouncements came from a group at Chi-ron. Chemists there developed a roboticsystem that can make millions of com-pounds in a few weeks using this ap-proach. The robot delivers chemicalsand performs the mixing and partition-ing of the solid support.

As we mentioned earlier, one of theproblems with a split-and-mix synthesisis identifying the composition of a reac-tive compound within a large mixture.Kit Lam of the University of Arizonahas developed a way to overcome thisobstacle. He noted that at the end of asplit-and-mix synthesis, all the mole-cules attached to a single bead are of thesame structure. Scientists can pull outfrom the mixture the beads that bearbiologically active molecules and then,using sensitive detection techniques, de-termine the molecular makeup of thecompound attached. Unfortunately, thistechnique will work only for certaincompounds, such as peptides or smallsegments of DNA.

Other investigators have developedmethods to add to each bead a chemi-cal label essentially listing the order inwhich specific building blocks have beenadded to the structure—in other words,the chemical equivalent of a UPC barcode. Reading the collection of theseso-called tags on a particular bead givesa unique signature and hence the identi-ty of the compound on that bead. Re-searchers at the biotechnology companyPharmacopeia, drawing on techniquesintroduced by W. Clark Still of Colum-bia University, have been very success-ful in applying powerful tagging tech-niques to their combinatorial libraries.

Nevertheless, because of the difficultiesof identifying compounds made in asplit-and-mix synthesis, most pharma-ceutical companies today continue torely on parallel synthesis.

Drug Libraries

Both the parallel and the split-and-mix techniques of combinatorial

chemistry began as ways to make pep-tides. Although these molecules are im-portant in biological systems, peptideshave limited utility as drugs because theydegrade in the gut, they cannot be ab-sorbed well through the stomach and


A2

A3

POLYSTYRENEBEAD

A1 MOLECULE

B2

B3

B1 MOLECULE

STEP 1Start with test tubes holding a solution con-taining inert polystyrene beads (gray circles).For simplicity, this example shows only threecontainers, but dozens might be used. Add thefirst set of molecules—the A class (squares)—to the test tubes, putting A1 molecules intothe first container, A2 molecules into the sec-ond, and so on.

STEP 2 Mix the contents of all the test tubes.

STEP 3 Split the mixture into equivalent portions.Then add the second set of molecules—the Bclass (triangles)—placing B1 molecules into thefirst test tube, B2 into the second, et cetera.(Repeat steps 2 and 3 as many times as need-ed, depending on the number of sets of build-ing blocks to be added.)

STEP 4Separate the beads from any unreacted chem-icals and detach the final structures. Research-ers often screen the contents of each test tubeto determine the mixture’s average biologicalactivity. Because each mixture shares the samefinal component, workers can determine whichvariant scores best—say, B2 might be most po-tent. They repeat the synthesis, adding only B2to the A compounds to find which of the A-B2combinations are the most biologically active.JA

RED

SC

HN

EID

MA

N D

ESIG

N

Split-and-Mix Synthesis


they are rapidly cleared from the blood-stream. The pharmaceutical industry be-gan to pursue combinatorial methodsmore aggressively after realizing thatthese techniques could also be applied todruglike compounds, such as the class ofmolecules known as benzodiazepines.

Benzodiazepines are one of the mostwidely prescribed classes of medicines.The best-known representative is diaze-pam, or Valium, but the class includes anumber of other derivatives with impor-tant biological activity: anticonvulsantand antihypnotic agents, antagonists ofplatelet-activating factor (a substanceimportant in blood clotting), inhibitorsof the enzyme reverse transcriptase inHIV, and inhibitors of Ras farnesyl trans-ferase (an enzyme involved in cancer).Because of the broad activity of this class,benzodiazepines were the first com-pounds to be studied in a combinatorialsynthesis seeking new drugs. In 1992one of us (Ellman), working with BarryBunin, also at the University of Califor-nia at Berkeley, described a way to syn-thesize benzodiazepines on a solid sup-port, making possible the synthesis oflibraries containing thousands of ben-zodiazepine derivatives.

More recently, the two of us (Plunkettand Ellman) worked out a better ap-proach for making benzodiazepines ona solid support; our new synthesis pro-vides easy access to much larger num-bers of compounds. The most challeng-ing aspect of any combinatorial synthe-sis is determining the experimentalconditions that will minimize side reac-tions giving rise to impurities. We spent

more than a year fine-tuning the reac-tion conditions for our new benzodi-azepine synthesis, but after determiningthe optimal procedure, we, along withBunin, generated 11,200 compounds intwo months using a parallel synthesis.

Promising Leads

From our benzodiazepine libraries,we have identified several com-

pounds with promising biological activ-ity. In a project with Victor Levin andRaymond Budde of the University ofTexas M. D. Anderson Cancer Centerin Houston, we have identified a benzo-diazepine derivative that inhibits an en-zyme implicated in colon can-cer and osteoporosis. And incollaboration with GaryGlick and his colleagues atthe University of Michigan,we discovered another ben-zodiazepine that inhibits theinteraction of antibodies withsingle-strand DNA—a pro-cess that may be involved insystemic lupus erythemato-sus. These compounds arestill in the very early stagesof laboratory testing.

Once we and others dem-onstrated that combinatorialchemistry could be used toassemble druglike molecules,the pharmaceutical industrybegan pursuing more proj-ects in this area. In the pastfive years, dozens of smallcompanies devoted entirely

to combinatorial chemistry have begunoperation. Nearly all the major pharma-ceutical companies now have their owncombinatorial chemistry departmentsor have entered a partnership with asmaller, specialized company that does.

As might be expected, researchershave branched out beyond benzodiaze-pines, routinely applying combinatorialtechniques to a wide array of startingmaterials. In general, chemists use com-binatorial libraries of small organicmolecules as sources of promising leadcompounds or to optimize the activityof a known lead. When searching for anew lead structure, researchers oftengenerate large libraries, with tens ofthousands or even millions of finalproducts. In contrast, a library designedto improve the potency and safety of anexisting lead is typically much smaller,with only a few hundred compounds.

Several pharmaceutical companies arenow conducting human clinical trials ofdrug candidates discovered throughcombinatorial chemistry. Because suchprograms are relatively new, none ofthese candidates has yet been studiedlong enough to receive approval fromthe U.S. Food and Drug Administra-tion. But it is only a matter of time be-fore a medicine developed with assis-tance from combinatorial methodsreaches the market.

Pfizer has one example in its pipeline.Using standard methods in 1993, thecompany discovered a lead compoundthat appeared to have potential for pre-venting atherosclerosis, or hardening ofthe arteries. In less than a year, using

Combinatorial Chemistry and New Drugs72 Scientific American April 1997

POLYSTYRENE BEADS (magnified roughly 100times) are often used in combinatorial chemistry sothat the products, which are attached to the beads,will be easy to separate from unreacted material.Beads that tested positive in a screening assay forartificial steroid receptors turned red.

SCREENING ROBOT at Arris Pharmaceutical automatically shuttles microtitreplates bearing combinatorial libraries to the equipment that tests for biological activity.

CO

UR

TESY

OF

W. C

LARK

STI

LL

CH

ARL

ES O

’REA

R


parallel synthesis, one laboratory at Pfi-zer generated in excess of 1,000 deriva-tives of the original structure, some ofwhich were 100 times more potent thanthe lead compound. A drug derivedfrom this series is now in humanclinical trials. Notably, work-ers made more than 900molecules before they no-ticed any improvement inbiological activity. Few labora-tories making standard one-at-a-time modifications to a lead compoundcould afford the time and money togenerate nearly 1,000 derivatives thatshowed no advantage over the originalsubstance.

Chemists at Eli Lilly also used paral-lel synthesis to develop a compoundnow in clinical trials for the treatmentof migraine headaches. They had earli-er found a lead substance that boundeffectively to a desired drug target, orreceptor. But the lead also had a highaffinity for other, related receptors, be-havior that could produce unwantedside effects. Researchers used parallelsynthesis to make approximately 500derivatives of that lead before arrivingat the one currently being evaluated.

Researchers will inevitably find waysto generate combinatorial libraries evenfaster and at a lower cost. Already theyare working out clever reaction meth-ods that will enhance the final yield ofproducts or replace the need for addingand later removing polystyrene beads.The future will also see changes in howinformation about the activity of testedcompounds is gathered and analyzed inthe pharmaceutical industry. For exam-ple, data on how thousands of com-pounds in one combinatorial library

bind to a partic-ular receptor can beused to predict the shape,size and electronic chargeproperties of that receptor, evenif its exact structure is unknown.Such information can guide chemists inmodifying existing leads or in choosingstarting materials for constructing newcombinatorial libraries.

Although the focus here has been thediscovery of drugs, the power of combi-natorial chemistry has begun to influ-ence other fields as well, such as materi-als science. Peter G. Schultz and his col-leagues at the University of California atBerkeley have used combinatorial meth-ods to identify high-temperature super-conductors. Other researchers have ap-plied combinatorial techniques to liq-uid crystals for flat-panel displays andmaterials for constructing thin-film bat-teries. Scientists working on these proj-ects hope to produce new materialsquickly and cheaply. Clearly, the full po-tential of this powerful approach is only

beginning to be realized.Combinatorial chem-

istry can appear somewhatrandom: combining various

building blocks and hoping some-thing useful comes out of the mix

may seem to be the triumph of blindluck over knowledge and careful pre-diction. Yet this impression is far fromthe truth. A good library is the result ofextensive development and planning.Chemists must decide what buildingblocks to combine and determine howto test the resulting structures for bio-logical activity. Combinatorial chem-istry allows researchers to gather, orga-nize and analyze large amounts of datain a variety of new and exciting ways.The principle of selecting the most ef-fective compounds from a collection ofrelated ones—the guiding axiom of theimmune system—is changing the waychemists discover new drugs. It is apleasant irony that this lesson gleanedfrom our natural defenses can be help-ful when those defenses fail.


The Authors

MATTHEW J. PLUNKETT and JONATHAN A. ELL-MAN worked together at the University of California,Berkeley, on combinatorial techniques for use with druglikemolecules. After receiving his Ph.D. from Berkeley in 1996,Plunkett moved to Arris Pharmaceutical, where he special-izes in making combinatorial libraries for random screeningand protease inhibition. Ellman joined the faculty of Berke-ley in 1992. His laboratory is currently engaged in the devel-opment of new chemistry for the synthesis of organic com-pound libraries and in the application of the library ap-proach to different problems in chemistry and biology.

Further Reading

Synthesis and Applications of Small Molecule Libraries. Lorin A.Thompson and Jonathan A. Ellman in Chemical Reviews, Vol. 96, No. 1,pages 555–600; January 1996.

Combinatorial Chemistry. Special Report in Chemical & EngineeringNews, Vol. 74, No. 7, pages 28–73; February 12, 1996.

Combinatorial Chemistry. Special Issue of Accounts of Chemical Re-search. Edited by Anthony W. Czarnik and Jonathan A. Ellman. Vol. 29,No. 3; March 1996.

High-Throughput Screening for Drug Discovery. James R. Broachand Jeremy Thorner in Nature, Vol. 384, Supplement, No. 6604, pages14–16; November 7, 1996.

SUPERCONDUCTORS can also be produced by combinatorialchemistry. This library of 128 copper oxide (CuO) compoundsyielded several superconducting films.

SA

CO

UR

TESY

OF

PETE

R G

. SC

HU

LTZ


combinatorial chemistry and new drugs

Documents