Figure 1.11. Peptides substituted with unusual carboxyl surrogate residues.
Figure 1.11. Peptides substituted with unusual carboxyl surrogate residues.
(117), and ureas (118)as well. In these libraries, the side-chains project from each fourth rather than each third atom in the chains so these are not close models of amino acids. Such libraries, not surprisingly, are more often antagonist rather than agonist.
In Fig. 1.11, one sees the insertion of unusual peptide bond surrogates for lead seeking. Such residues include peptide boronates, peptide hydroxamates, peptide aldehydes, peptide trifluoromethylketone hydrates, peptide nitriles, peptide phosphonates, and so on. One could also add to this list inclusion of jS-turn mimics, /3-sheet analogs, and so on. This is at present a very active subfield of medicinal chemistry.
These peptide and pseudopeptide libraries have been replaced progressively by collections containing smaller and more drug-like molecules. These will be covered in their own section below.
The minimum fully realized library of natural peptides would consist of 202° components. An analogous library of nucleic acids would consist of 55 components (double this if one used both ribose and deoxyribose units) and be substantially smaller (Fig. 1.12a). Once again, the use of post-translationally modified bases or wholly unnatural analogs increases the attainable diversity. Conformational effects and self-associations further enhance the diversity. Once again, construction is iterative and bead-based automated procedures are available. Libraries of significant size have been constructed and evaluated (119).
Construction of diverse oligosaccharidelibrar-ies is much more difficult. The linkage is chiral and relatively hard to control, the bonds are acid fragile, and there are many potentially competing functional groups that can be points of attachment (Fig. 1.12b). Despite these complicating factors, such libraries are beginning to appear. Clearly progress is being made.
3.4 Lipid Arrays
Lipid libraries have largely been neglected. For saturated fatty acids, the construction of carbon-carbon bonds is more difficult. Ste-
roids and other polyisoprenoids are also complex to construct. Mixed triglycerides would seem accessible, and one anticipates developments in this area.
4 SOLID AND SOLUTION PHASE LIBRARIES OF SMALL, DRUCABLE MOLECULES
As noted above, the field of combinatorial chemistry and multiple parallel synthesis started with libraries of peptides. In time, unusual residues crept into the products. While this evolution is still ongoing, it is now accompanied by a major effort to produce libraries of small, drug-like molecules in library form. Many of the methods used for large molecules carry over but the largely non-iterative nature of small molecule synthesis is a significant complication.
The current "gold standard" in small molecule drug seeking is oral activity accompa nied by one-a-day dosing. This is a high hurdle. The majority of molecules that have passed have molecular weights of about 500 or so. It has been calculated that the number of small molecules that would fall into this category is approximately 10 raised to the 62nd power! Clearly preparing all of these in reasonable time is beyond the capacity of the entire population of the earth even if they worked tirelessly. The number of compounds that can become satisfactory drugs encompassed in this impossible collection is probably in the range of a few thousand, so most of the effort would be wasted. Hence medicinal chemical skills are still at a premium.
Obviously construction one at a time in the usual iterative or non-iterative fashion results in single molecules of a defined nature. Combination of reactants A and B produces a single product A-B. Reaction of this with another substance produces product A-B-C. In each case, a single reaction produces a single product. The change brought about by combinatorial or multiple parallel synthesis methods is that the reactants are usually linear, but the products can be logarithmic. For example, reacting A with 10 different Bs produces 10 products (A-B^o), and the reaction with 10 different Cs on each of these results in 100 products (A-B^qCho), either in mixtures or as discrete compounds. Rather large compound collections can be assembled quickly using this scheme.
With non-linear products a different variant is seen than experienced with large molecules (Fig. 1.13). Here a starting material (often called a centroid) with a number of functional groups (preferably with different degrees of reactivity-known as orthogonality) can be reacted with a variety of substitu-ents (often called adornments) to produce a large number of analogs. In the figure, one sees illustrated centroids with two, three, or four such functional groups and given the number of possible variants, this can lead to a very large library of analogs in a brief time (two functional groups with 10 variant adornments quickly results in 100 analogs, three in 1000, and four in 10,000). If the reaction conditions allow, these variations can be run in mixture or in parallel effecting a very significant time savings. It can, however, put a sig-
Two substitutable groups
Three substitutable groups
Four substitutable groups c aG)b d
Figure 1.13. Centroid adornment.
nificant strain on purification, analysis, record keeping, budgets, etc. Much work has been expended in addressing these potential limitations.
With 500 as the normal practical upper molecular weight limit, it can be seen that centroid A should be chosen to have the smallest practical molecular weight. It is also helpful if, when fully adorned, it has functional groups remaining that can interact productively with a receptor so the weight devoted to this part of the molecule is not net loss. The molecular weight of the centroid places practical limits on the net weight that the adornments can collectively have. If one adornment is rather large, then this requires one or more of the other adornments to be made compensatingly smaller. The more functional groups present in the centroid, the smaller each adornment can be.
It is particularly helpful if the adornments project into space into quadrants that fit precisely the needs of the receptor if one is optimizing a lead. Alternately, if one is hit seeking, they should project into various quadrants about the centroid so as to allow a fruitful exploration of potential receptor needs. In hit seeking, one often wants molecular flexibility, whereas in lead optimization progressive ri-gidification is often more effective.
In addition, the adornments must have the usual medicinal chemical characteristics. They should not be chemically reactive, convey toxicity, or be inordinately polar. The product should fit the modified Lipinski rules to allow for the usual molecular weight and lipophilicity creep that often accompanies analoging. The net hydrogen bonding inventories, log P, water solubility, and cell penetrability features set constraints on the individual and the collective nature of the adornments. If one is rather polar, for example, the polarity of another usually must be decreased. Whether the centroid should be tethered to a solid support or free in solution must be considered carefully. If tethered, it is important to consider whether the point of attachment will remain in the final analog, and if so, what affect this may have on its biological properties. One also should consider whether this attachment will prevent the use of one of the potential adornment points.
Just as in one at a time synthesis, linear syntheses are the most risky and produce the lowest yields. Converging methodologies address these limitations successfully, and in combinatorial work, Ugi (four-component) and Passerini (three-component) reactions are very flexible and popular. Generally one has less control over the specific products being produced by such reactions but this is largely compensated for by the molecular diversity available in this way.
Clearly a great deal of thought should go into library design before the work begins.
The first libraries containing heterocycles recognizable as orally active drugs were the prepared on resins by Bunin and Ellman in 1992 (Fig. 1.14) (6) A notable chemical feature is the use of amino acid fluorides to drive the amide formation to completion. The choice of benzodiazepines was inspired because of the medicinal importance of these materials and their resemblance to peptides. Here the library was constructed by a combination of three reactants. If each were represented by 10 variations, the library could easily reach 1000 members (10 X 10 X 10) in short order. This would fit the commonly accepted meaning of combinatorial in that all of the possible variants would be constructed. Being selective in the variations actually incorporated, a smaller ("focused") library could be made that answered specific pharmacological questions but at the risk of missing an unexpected discovery. Such a li-
i. N-alkylation ii. Detach
Figure 1.14. Synthesis of benzodiazepine libraries—1.
brary would fit the commonly accepted definition of multiple parallel synthesis. Such focused libraries are often more intellectually satisfying to the practitioner. In this library the attached groups project into space at widely separated compass points around the molecule allowing a systematic exploration of receptor requirements. The centroid has a molecular weight of 160 when all of the available substitution points are occupied by hydrogen atoms. If one accepts an upper molecular weight limit of approximately 500, then four variations can occupy 340 atomic mass units (500-160) so each adornment can have an average of about 85 amu, if the weight is evenly distributed. This gives significant latitude for substitution. When chosen with care to convey drug-like properties and not to exceed collectively the guidelines that Lipinski has developed, the library can contain primarily substances that have a chance to be drug. They also can be so chosen that they allow for sub stitution independently of each other and also to be installed without premature separation from the beads. It will also be noted that the centroid chosen has amide and amine linkages that are not involved in adornment attachment and are capable in principle of interacting successfully with a receptor so the molecular weight sacrificed to the centroid may perform pharmacodynamic work. Centroids derived from molecular series that are known to be associated with good pharmacokinetics are often referred to as privileged molecules. Thus, the choice of benzodiazepines to demonstrate the potential power of combinatorial chemistry and multiple parallel synthesis was inspired.
In this pioneering library, the final products were attached to the bead support through a phenolic hydroxyl group that remained as such in the products before testing. Varying the point of attachment of the hy-droxyl group would lead to additional multiple
Figure 1.15. Synthesis libraries—2.
analogs. Indeed, a variant of this process resulted in a traceable linker in the other aromatic ring (Fig. 1.15).
A somewhat more versatile synthesis of this type using stannanes and palladium acy-lations (Stille coupling) appeared subsequently (Fig. 1.16) (120). While precedent establishing, this was pharmacologically less than completely satisfying because agents intended to penetrate well into the CNS should not usually contain such a polar substituent. Despite this, several components in these libraries were bioactive, and the work drew widespread attention to the promise of the methodology and was soon followed by a flood of applications to the preparation of drug-like molecules. In this sequence, attachment to the resin was by an amino acid ester bond. Subsequently this bound intermediate was converted to an imine that cyclized to the benzodiazepine moiety on acid cleavage from the resin (Fig. 1.18) (121). The "traceless linker" technology so introduced has now become standard. One of the additional advantages of this application is that incomplete reaction occurring during the synthesis would lead to products that would not cleave from the resin and could be removed by simple filtration. Furthermore, the products were now indistinguishable from benzodiazepines prepared by usual speed analoging (USA) methods.
Ellman's group also developed a traceless linker sequence of a different type based on HF release of an aryl silicon link to the resin (Fig. 1.17) (122).
As it happens, somewhat gratifyingly, testing of these agents revealed no structural si^r-prises. The intense study of the benzodiaz-epines in the empirical earlier years had apparently not missed much of significance. Nonetheless, these studies resulted in con-
Was this article helpful?