What Is Combinatorial Chemistry? There are three common features that describe a combinato-
rial chemistry project (reviewed in refs. 2–9). The first is the
type of molecules that comprise the library itself. Combinato-
rial libraries have been described that are composed of com-
pletely random sequences of peptides or oligonucleotides.
Libraries have also been described that consist of random,
site-directed mutants of a specific protein or nucleic acid
oligonucleotide, and are therefore composed of many variants
of an initial parental molecule. Finally, combinatorial libraries
of small organic molecules can be generated by a variety of
synthetic methods, leading to the synthesis and screening of a
family of specific small molecules for potential utility as a drug.
In any combinatorial library, regardless of the type of
molecules represented, all of the compounds are related to one
another. Their structures are all built from a common set of
chemical building blocks, with each molecule possessing a
unique combination or sequence of these building blocks at
each synthetically incorporated position. Additionally, the
molecules all possess a common structural core or synthetic
linkage, dictated by the type of molecules in the library and by
the actual synthetic strategy employed. For example, collec-
tions of peptides or protein molecules in a combinatorial
library are usually built from the 20 naturally occurring amino
acids, and possess a common synthetic linkage (an amide
bond) between each position in the polymeric molecule.
The second feature of a combinatorial experiment is the
diversity that can be experimentally attained and exploited.
Any library that can be encoded genetically is potentially
capable of containing hundreds of millions of separate, related
molecules. For example, the second talk of this session (Wells)
described the screening of over 10
7
mutated variants of the
human growth hormone (hGH), using recombinant DNA
methods to screen each separate molecule on the surface of a
unique viral clone. Because any one clone contains, in a single
viral package, expressed copies of the actual molecule of
interest and the genetic sequence encoding that molecule, the
recovery of a single copy of a useful construct allows the
determination of the precise sequence and structure of that
molecule.
In contrast, combinatorial experiments that rely on the
manual chemical synthesis of individual molecules face a more
serious problem of attainable and useful diversity, as described
by Jon Ellman. Unlike genetic combinatorial methods that
Copyright
q
1997 by T
HE
N
ATIONAL
A
CADEMY OF
S
CIENCES OF THE
USA
0027-8424
y
97
y
942779-4$2.00
y
0
PNAS is available online at http: yy