Synthetic combinatorial methods, combined with rapid assays, have fundamentally advanced the ability to synthesize and screen large numbers of compounds. Using a range of combinatorial approaches, libraries composed of tens of thousands to millions of different compounds have been produced. Combinatorial chemistry and high-throughput screening is now a universally utilized tool for drug discovery and development, but the harsh reality is that the drug discovery process remains extremely slow and enormously expensive. Drug candidates resulting from many combinatorial approaches have also often tended not to have drug-like properties and thus have a high inherent rate of attrition in the later stages of drug development because of poor physicochemical properties. Although unrelated to the advances in combinatorial approaches, it is worth noting that increased regulatory issues and unrealistic public expectations have reduced the number of approved drug entities over the past 20 years from approximately 35 per year to 10 or less.

One approach to circumvent this high attrition rate would be to use in vivo models directly in the discovery phase to identify candidates with desired biological profiles while simultaneously eliminating those compounds with poor absorption, distribution, metabolism, and elimination (ADME)/pharmacokinetic (PK) properties.

It is clearly unrealistic to use discovery in vivo models to screen the large collections of hundreds of thousands of the individual compounds currently available. A potential solution that shows promise is the use of mixture-based combinatorial libraries directly for in vivo testing. This offers a unique opportunity to carry out successful preliminary studies in which tens to hundreds of thousands of compounds would be screened directly in translational in vivo assays. This has been accomplished in early studies carried out in rats and dogs to monitor blood pressure and heart rate (Houghten, 1994) using 400 separate mixtures each of 132 000 hexapeptides. Immunological modulation by large mixtures has also been accomplished (Shukaliak Quandt et al, 2004). Recent studies have involved research into pain therapeutics utilizing in vivo models with a tetrapeptide library made up of a total of 6 250 000 peptides (200 mixtures made up of 125 000 tetrapeptides each) (Dooley et al, 1998; Houghten et al, 2006, 2008). Mixtures ranging from 2500 to 125 000 tetrapeptides have yielded clear in vivo activity that is not necessarily related to classic in vitro target-based screening. For mixture-based small molecule libraries the process can be improved by careful selection of those libraries guided by theoretical calculation of their drug-like properties. Over the past 10 years a process termed cassette testing (Liu et al, 2008, and references cited therein) has been used to study in vivo ADME with small mixture sets (typically 5–10 related compounds) to facilitate the early elimination of compounds with poor drug-like profiling in PK profiling.

The concept of using large, highly diverse mixture-based libraries for the identification of inherently more advanced ‘hits’ by the direct in vivo testing is both exciting and promising. It remains to be seen if these recent early preliminary successes will fulfill their current potential promise.