Encoding strategies in combinatorial chemistry

To Pasteur’s famous dictum, “Chance favors the prepared mind,” could arguably be added the corollary, “As the number of chances increases, so increases the effect of favor.” It is nearly a truism that a great many important discoveries in chemistry originated not from careful design, but from astute observations of unintended experiments. Suppose, then, that creativity were applied to the task of enormously increasing both the number of “chance” experiments possible and one’s confidence that a “success” would be observable? In essence, that is the philosophy behind the new field called combinatorial chemistry: Design a synthesis experiment that both leads to a large collection of new chemical substances (“molecular diversity”) and affords that collection in a format facilitating the selection of individual members possessing a given property of interest.

This paper is organized on our German–American Frontiers of Science session on “Combinatorial Chemistry,” which was held on June 19, 1996, in Munich. The presenters were Michael Famulok of the Ludwig-Maximilians-Universität, Jon Ellman of the University of California at Berkeley, and the author. J. Ellman has previously published a summary paper in the PNAS on “Combinatorial thinking in chemistry and biology” (1), and so I will not repeat that topic. Instead, I found it of interest that each presenter’s combinatorial chemistry experiments embodied fundamentally different types of encoding strategies. Thus, this paper will summarize the proceedings as a vehicle highlighting the concept of encoding and surveying the merits of different encoding methods.

What Is Encoding?

A combinatorial chemistry or biology approach to discovery begins with a large collection of unique substances, often molecules, which is the source of molecular diversity. By the process of screening, the members of this library that have “desirable properties” (as determined by the screen) are culled out. It now becomes essential to learn the chemical identity of each “winning” library member. You cannot tell this simply by “looking” at the winners, and there is no percentage in having a “winning” molecule sitting in a vial if you cannot determine what it is. The various methods of establishing the identity of the winners constitute the set of useful encoding strategies.

Positional Encoding.

J. Ellman described the synthesis of various organic compound libraries. In his work, the solid-phase support for synthesis begins with a one-piece polypropylene rack of pins arrayed in a 12 × 8 matrix. The synthetic method is designed so that each pin bears a different, single organic compound (assuming the synthetic method worked). Because the arrangement of each pin in the rack cannot change, the physical position serves as the code for each library member. That is, after a screen is complete one might find that the compound on pin E-7 was a “winner.” By checking back on how the synthesis was conducted, one can know the chemical identity of the compound on pin E-7. Because it is an assumption that all reactions worked, the very first next step is to resynthesize the compound you expect to be on pin E-7, to thoroughly characterize that compound, and to verify that it indeed is a “winner.”

The method of positional encoding was utilized in the first synthetic combinatorial library, in which peptides were synthesized on such a rack of pins and screened for antibody binding ability (2). Such a two-dimensional matrix approach is also the encoding method currently in use for the “DNA chip” technologies.

Chemical Encoding.

Using the in vitro selection method, M. Famulok described the generation of vast numbers (billions and up) of RNA sequences in a small volume of solution. The solution was passed through an affinity column, and RNA sequences that bound tightly to the immobilized ligands eluted more slowly than others. By necessity, only minuscule amounts of each sequence are present in the starting mixture. Polymerase chain reaction (PCR) technology permits the amplification of such minuscule amounts of nucleic acid sequences. After the amount of RNA available for analysis has been amplified, the RNA sequence can be determined by using existing analytical methods. By establishing the RNA sequence of the “winner,” its chemical identity becomes known. While this thinking may sound circuitous, the determination of chemical identity may also be thought of as simply the most direct method of chemical encoding. If the “winner” of a screening experiment had been 10 mg of a small organic molecule, its structure could have been determined by using standard characterization methods (NMR, mass spectrometry, microanalysis, etc.) and the sample would have encoded its own identity by way of direct chemical determination.

Of course, it is unlikely that any “winner” from a synthesized library would ever be obtained in 10-mg amount. (A 10,000-member library would have to contain 100 g, and that would be very hard to screen indeed.) While it is possible to synthesize mixtures of small molecules each in a tiny amount, determining the identity of “winners” is accomplished by a process of deconvolution rather than encoding. [This is because the only code in such a library is chemical identity, and at such tiny amount the only tool useful to establish identity is mass spectrometry (3–8).] Instead, such small molecule libraries are produced on polymer beads, each bead possessing only one type of molecule (9). Encoding of each bead can then be accomplished with a different type of chemical encoding, a chemical tag. The first chemical tags were nucleic acids, because their sequence can be established with an exceedingly tiny amount of the tag (10). Rather recently, more robust forms of chemical tags based on abiotic molecules have been described (11).

Electronic Encoding.

My own current work is in the area of electronic encoding (12). This form of encoding is today accomplished by using a microelectronic device called a radiofrequency (rf) memory tag (13, 14). The tag, measuring 13 × 3 mm, is encased in heavy-walled glass and contains the following: (i) a silicon chip, onto which is laser-etched a unique binary code (with 2⁴⁰ possible codes, no two chips need ever be manufactured with the same code); (ii) a rectifying circuit, with which absorbed rf energy is converted to D.C. electrical energy; (iii) a transmitter/receiver circuit; and (iv) an antenna, through which energy is received and rf signals are both received and sent. No external power source is required. Placing this chip near a computer-interfaced transceiver causes the chip to power up, read its identification tag, and send that code to the transceiver in an rf pulse. A reaction platform (e.g., a bead, tube, bag, or can) bearing an rf chip is thus uniquely tagged in a machine-readable fashion. Automated handling is facilitated by the near nondirectionality of the reading frame. A sorting machine designed to send rf-tagged reaction platforms to vessels for reaction or cleavage has been described (12).

Other Encoding Methods.

Several encoding strategies have been described whose use was not exemplified in this session. These include graphical (15, 16, †), optical (17, 18, ‡), NMR-based (19), and physical (20) methods of encoding. All are currently in much lower usage than the three strategies described above.