This paper reviews some of the broad issues associated with the production and screening of combinatorial libraries and proposes a guideline for optimizing the utility of combinatorial chemistry in drug discovery. This guideline is based on the premise that our knowledge of how diseases, biomolecular targets, and biologically active compound classes interrelate can be used to define the most productive regions of molecular diversity space. Compound classes known to modulate function in various disease-related biomolecular target classes provide rich, validated pharmacophores and should be given highest priority in the design and construction of combinatorial libraries. This selection system is illustrated with alpha-ketoamide libraries for the inhibition of serine and cysteine proteases and with oxindole libraries for the inhibition of protein kinases.

The considerable gain in efficiency achievable by introducing the principle of convergence into automated parallel synthesis is illustrated with a seven-step synthesis of a sixteen-hundred-compound array of potential serine protease inhibitors, conducted on a 50 mu mole scale (ca. 20 mg. of each compound).