CHEMISTS everywhere are losing their patience. No longer are they content to
spend years painstakingly assembling a designer molecule, only to find that
their new chemical fails to make the grade as the wonder drug or revolutionary
catalyst they were hoping for. So instead of taking aim at a single molecular
structure, they are turning to a scatter-gun approach and making hundreds or
thousands of different compounds all in one go. That way, they reason, they can
only improve their chances of finding the one with the magic properties
they are
seeking.
The name of the new game is combinatorial chemistry. Chemists take a number
of molecules as building blocks and join them in all possible combinations. The
goal is to make as many different molecules as possible, while still keeping
track of exactly how each molecule was put together. The result is a 鈥渓ibrary鈥�
of maybe tens of thousands of different chemicals that can be tested for
whatever properties are of interest.
It鈥檚 as if鈥攁fter a particularly bad darts night in the pub鈥攜ou
decided to design a better dart. Traditionally, you鈥檇 look at the best
darts and
consider what would make them fly more accurately. Then you鈥檇 choose the point,
body and feathers accordingly. The combinatorial alternative is to take as many
dart parts as you can, assemble them in all possible combinations and then aim
all the darts at the board. You can then pick out the one that works best and
examine it to find out how it was put together.
Advertisement
The most straightforward technique in combinatorial chemistry is known as
鈥減arallel synthesis鈥�. Take the simple case of a common reaction in which an
amine reacts with one of a class of chemicals called acyl chlorides. With five
amines reacting with five acyl chlorides, it is possible to create 25 different
reaction products simply by making all possible combinations in 25 test tubes.
From these products, you could make an even more diverse array of chemicals.
Simply divide each of the 25 products into four test tubes and carry out
reactions on them using four new reagents to give 100 different products.
Today, there are robots that can automate parallel synthesis, adding
predetermined combinations of reagents to each tube. But it is not hard to see
that the number of test tubes you need starts to rocket if you are planning a
sequence of several reactions, using even a modest number of alternatives at
each stage. In practice, this means that parallel synthesis cannot be used to
produce more than about 100 compounds at a time.
Taking off
The newer technique of 鈥渟plit synthesis鈥�, which can generate much larger
libraries, has really allowed combinatorial chemistry to take off. Split
synthesis has its roots in the 1960s, when Bruce Merrifield of the Rockefeller
Institute for Medical Research (now Rockefeller University) developed a method
of automating the synthesis of the long sequences of amino acids known as
peptides. That work won Merrifield the 1984 Nobel Prize for Chemistry.
The key trick, from the modern standpoint, was not the automation itself but
Merrifield鈥檚 ability to grow each peptide on a small, chemically inert polymer
bead. Previously, making a 10-amino-acid peptide, say, would have meant working
in solution. As each of the 10 steps was completed, the chemist would have had
to isolate and purify the growing compound. But the solid polymer beads could
simply be filtered out at each stage, with the growing peptide attached.
Split synthesis uses similar polymer beads, allied to the combinatorial
approach (see
Diagram, p 24). You could take 100 000
polymer beads, typically
about 10 micrometres across, and attach the same starting molecule to
each. Then
divide the beads into, say, five groups and perform a different chemical
reaction on each. Now there are five different compounds. Mix all the beads up,
then divide them again, so that you have five new groups each with an equal
number of beads from the first groups. If you now perform another five
reactions, you end up with 25 compounds. Repeat the process four more times and
you鈥檒l have a staggering 15 625 different compounds, with each molecule
attached
to its own bead.FIG-20384001.gif
This all sounds simple enough. The hard part is deciding which building
blocks to use, and which reactions to apply to them. 鈥淪ome people have the
conception that it鈥檚 just brute force,鈥� says Jonathan Ellman, a pioneer of
combinatorial chemistry at the University of California, Berkeley. 鈥淏ut
designing that chemistry and being successful at producing those
compounds鈥攊t takes thought.鈥� The chemical building blocks and reactions
have to be carefully selected to create a rich source of diverse molecules
rather than vast numbers of fairly similar products. These reactions must also
be thoroughly tried and tested, so that they are guaranteed to work at every
stage in the process.
Another essential step is to tag the molecules as they grow, so that
each can
be identified at the end. Several techniques are now available to do this. In
one, each step of the process adds an inert compound to a separate site on the
bead, creating a chemical signature that can later be read using gas
chromatography (New 杏吧原创, Science, 8 July 1995, p 18). In
another,
each stage adds a nucleotide, one of the four 鈥渓etters鈥� in the DNA alphabet. The
sequence can later be read using a technique known as polymerase chain reaction
(Technology, 29 January 1994, p 17).
Memory beads
Researchers in La Jolla, California, have even devised a way of tagging the
molecules using microchip memories. Last year, K. C. Nicolaou of the Scripps
Research Institute, Xiao-Yi Xiao of the chemicals company IRORI Quantum
Microchemistry and their colleagues grew molecules on beads that had been
packaged up in a capsule, along with a microchip. The walls of the
capsules were
made of a mesh that lets the chemical reagents through, but was fine enough to
prevent the polymer beads escaping. At each stage, the researchers sent a radio
signal to each chip to record details of the reaction the bead was exposed to.
The details could be read off the chips when the molecules were complete.
One big advantage of the electronic tag is that it can be read in an
instant.
鈥淵ou don鈥檛 need fancy analytical techniques to decode it,鈥� says Xiao.
Electronic
tagging also allows chemists more freedom in the molecule-building reactions
they can choose: they do not have to avoid those that destroy the often fragile
chemical tags. But the technique is still in its early stages, and has some
drawbacks too. Each microchip needs its own communications aerial, and so is
fairly big鈥攁bout 12 millimetres long. This makes the microchip system too
bulky to be used for producing large libraries of hundreds of thousands of
compounds. Xiao and his colleagues are trying to cut down the size of their
high-tech beads, and they hope to produce a commercial tagging system soon.
In the pharmaceuticals industry, combinatorial chemistry is revolutionising
the search for new drugs. Stephen Kaldor, head of combinatorial chemistry
at Eli
Lilly in Indianapolis, Indiana, says that it is used in more than 75 per
cent of
the firm鈥檚 drug discovery programmes. Though the company has yet to market a
drug found in this way, Kaldor says that last November it began human toxicity
trials of a promising compound, found by combinatorial methods, that
targets the
nervous system.
Lilly found the drug through a process called 鈥渓ead optimisation鈥�, which
starts with a molecule that is already known to have the desired property.
Chemists then tinker with it to produce a number of molecules that have similar
structures, in hope that some of these relatives will have the same property,
only more so.
Other researchers prefer to take the opposite approach, and start with a
chemical structure no one yet knows much about. This approach, known as 鈥渓ead
generation鈥�, is particularly interesting to chemists because it explores truly
unknown territory. It has commercial advantages too, says Ellman, as it is less
likely that anyone will have previously patented any compounds that emerge.
Eric Jacobsen and his colleagues at Harvard University have taken the lead
generation approach to explore another area of chemistry鈥攖he search for
stereoselective catalysts. These direct the way in which one or other of a pair
of 鈥渟tereoisomers鈥�, mirror-image pairs of molecules, emerge from certain
reactions. When used as drugs, these mirror-image twins can have a dramatically
different effects. A notorious example is thalidomide: one stereoisomer causes
deformities in unborn children, while the other is a harmless sedative.
The drug
that pregnant women were prescribed back in the 1960s was a mixture of the two,
with devastating consequences. For substances like these, a catalyst capable of
forcing a reaction to produce only the desirable molecule would be hot
property.
Yet 20 years of study by traditional methods have yielded fewer than 10
stereoselective catalysts. From this small sample chemists have not been
able to
gain much of an insight into how such catalysts work, and so there has been
little hope of designing new ones in the traditional way. This has led many
researchers to see combinatorial chemistry as the way forward. For Jacobsen, it
amounts to a tool for seeking out serendipity. 鈥淐hance can play an important
role, and we want to maximise our chances,鈥� he says. 鈥淲e鈥檇 like to discover
catalysts or catalytic structures that we never would have imagined.鈥�
Jacobsen has now discovered stereoselective catalysts for three reactions
within his libraries of 10 000 or more compounds, but progress has been slow.
The problem is that testing the myriad compounds that combinatorial methods
produce can be very time-consuming. 鈥淵ou could take each bead and test it, but
if you had 10 000 beads, you might have to run 10 000 reactions,鈥� says
Jacobsen.
Even if each test took no more than a few hours, the process could take
years.
But there are hopes that this bottleneck could soon be cleared, as
techniques
for quickly testing huge numbers of certain compounds come onto the scene. For
instance, in 1991 Kit Lam of the Arizona Cancer Center in Tucson, Arizona,
showed that millions of peptides could be screened in three or four hours,
using
a test that stained only those beads bearing active compounds. Lam and his
colleagues created a probe by taking an antibody to the natural peptide
beta-endorphin and linking it to an enzyme that causes staining when the
antibody couples to a particular amino-acid sequence:
tyrosine-glycine-glycine-phenylalanine-leucine. The researchers then mixed the
probe with a library of nearly 2.5 million different five-amino-acid peptides,
which they had prepared using split synthesis on polymer beads.
Somewhere in their library, Lam believed, that sequence would have
formed. He
would be able to pick it out, because it would bind to the antibody-enzyme
target, causing its attached bead to turn turquoise. Sure enough, six
beads were
stained. Lam thinks that similar methods should work in any split synthesis
experiment. 鈥淭his method is potentially very powerful,鈥� he says.
Bottleneck
Jacobsen and his team have also been working to relieve the testing
bottleneck. He claims they have found a way to test many thousands of
stereoselective catalysts at the same time. They submitted a report on their
work to the Journal of the American Chemical Society in mid-May,
but it
has yet to be published and Jacobsen is reluctant to go into details.
Combinatorial chemistry also looks set to speed up the search for
superconductors and other new industrial materials. Superconductors, which have
zero electrical resistance, could revolutionise the electricity industry by
allowing power to be distributed without any loss en route. Unfortunately, most
materials that behave as superconductors only have this property at
temperatures
close to absolute zero. Of the so-called 鈥渉igh-temperature鈥� superconductors,
which have no resistance up to temperatures of around 鈭�140 掳C,
only a handful have been found, despite painstaking research round the
world.
Combinatorial methods produced no less than four high-temperature
superconductors in a single experiment in 1994. Researchers led by Xiao-Dong
Xiang of the Lawrence Berkeley National Laboratory and Peter Schultz of the
University of California at Berkeley made a library of 128 compounds by
depositing seven elements one after the other on inch-square wafers of
magnesium
oxide (MgO) or lanthanum aluminate (LaAlO3), which was masked so that
each element was blocked from part of the wafer. A different masking
pattern was
used for each element, and the patterns were chosen so that every possible
combination of the elements was found somewhere on the wafer. When the
deposition was complete, the wafers were heat-treated to force the elements to
react together (Science, vol 268, p 1738).
Because the family of superconductors based on copper oxide had already been
fairly firmly pinned down, none of the superconductors made by Xiang and
Schultz
was new. What their experiment proved, however, was that this method can cover
ground quickly. 鈥淲e chose seven elements that were likely to form copper oxide
compounds, based on their crystal structure, and we found all the
high-temperature copper oxide superconductors that people had discovered among
the combinations of these elements,鈥� says Xiang. 鈥淭his experiment was like a
proof-of-concept.鈥� He hopes to apply similar methods to other materials to find
new superconductors.
Xiang and Schultz have also adapted their technique to find new materials
whose electrical resistance changes in a magnetic field, a property called
magnetoresistance. These materials are already used in a variety of devices,
including the heads that read and write data to computer discs.
Researchers hope
to find materials that show bigger resistance changes, as this would lead to
disc drives that hold much more data (see 鈥淕iants in their field,鈥� 10 February,
p 34).
In 1993, large magnetoresistance effects started coming to light in
compounds
based on manganese oxide (MnO3). So Xiang and his team tried ringing
the changes: 鈥淲e thought, instead of manganese oxide, why don鈥檛 we try other
transition metals, which we know will probably have very similar properties.鈥�
Quite quickly, they hit the jackpot with cobalt oxide (Science, vol
270, p 273, 1995). 鈥淲e were pretty lucky,鈥� Xiang says. 鈥淲e just made a
couple of
them in this cobalt-based library and we found many new compounds that have
尘补驳苍别迟辞谤别蝉颈蝉迟补苍肠别.鈥�
Xiang is developing an automated process, which by late this summer could be
creating 10 000 compounds each day. But the bottleneck will once again be the
screening. Until now, the team has screened the new compounds by hand, by
placing them in different magnetic fields and applying a voltage across them.
But when new compounds start arriving in their thousands, this method will not
be able to cope.
So the researchers will analyse their new materials by imaging the surface o
f the library, using a scanning microwave microscope. Materials with high
resistance tend to absorb more of the microwave energy and reflect less. Xiang
expects the microwave images to measure the resistance of compounds that are
just 0.1 millimetres apart, and the system should allow 10 000 compounds to be
screened in just a few hours.
The future for combinatorial chemistry looks bright. Bigger and better
libraries, perhaps containing a million chemicals or more, are already in
sight.
The challenge now is to find testing methods that are fast enough to keep up.
One day, finding the sought-after molecules may be as easy as pulling winning
darts from a dartboard. What researchers can be sure of is that at their
chemical scatter-gun will hugely improve their chances of hitting the bull鈥檚
eye.