Combinatorial synthesis on solid phase can generate very large numbers of products, using a method described as mix and split synthesis. This technique was pioneered by Furka and has been enthusiastically exploited by many others since its first disclosure. For example, Houghten has used mix and split on a macro scale in a "tea bag" approach for the generation of large libraries of peptides.
The method works as follows: a sample of resin support material is divided into a number of equal portions (x) and each of these are individually reacted with a single different reagent. After completion of the reactions, and subsequent washing to remove excess reagents, the individual
portions are recombined, the whole is thoroughly mixed, and may then be divided again into portions. Reaction with a further set of activated reagents gives the complete set of possible dimeric: units as mixtures and this whole process may then be repeated as necessary (for a total of n times). The number of compounds obtained arises from the geometric increase in potential products; in this case x to the power of n.
A simple example of a 3 x 3 x 3 library gives all 27 possible combinations of trimeric products. X, Y and Z could be amino acids, in which case the final products would be tripeptides, but more generally they could be any type of monomeric unit or chemical precursor. It can be seen that the mix and split procedure finally gives three mixtures each consisting of nine compounds each, and there are several ways of progressing these compounds
to biological screening. Although the compounds can be tested whilst still attached to the bead, a favoured method is to test the compounds as a mixture following cleavage from the solid phase. Activity in any given mixture reveals the partial structure of active compounds within the library, as the residue coupled last (usually the N-terminal residue) is unique to each mixture. Identification of the most active compound relies
on deconvoluting the active mixtures in the library through further synthesis and screening.
In the example where the active structure is YXY, the mixture with Y at the terminal position will appear as the most active. Having retained samples of the intermediate dimers on resin (so-called "recursive" deconvolution ) addition of Y to each of the three mixtures will give all nine compounds with Y at the terminal position, and the second position defined by the mixture. The most active mixture here defines the middle position of the most active trimer to be residue X. Finally, the three individual compounds can be independently resynthesised and tested to reveal both
the most potent compound and also some structure activity relationship data.
In contrast, Lam et al. tested a family of peptides whilst still attached to the resin bead solid phase. Nineteen amino, acids were incorporated into pentapeptides to generate a library of almost two and a half million compounds. By using a colourimetric assay, beads bearing peptide sequences that bound tightly to the protein streptavidin or to an antibody raised against β-endorphin were revealed by visual inspection. Bead picking using micromanipulation isolated the beads, and the active peptide structures were determined by microsequencing.
A modification of this method has allowed screening of such libraries in solution. Linkers have been devised that allow several copies of the library compounds to be released sequentially. Using this method it is possible to identify an active mixture using a solution assay, and then return to the beads that produced these compounds, and redistribute them into smaller mixtures for retest. By repeatedly reducing the mixture size, ultimately to single compounds, the bead containing the most potent sequence may be identified and the peptide product sequenced.
