Designing combinatorial libraries optimized on multiple objectives.

Abstract

The recent emphasis in combinatorial library design has shifted from the design of very large diverse libraries to the design of smaller more focused libraries. Typically the aim is to incorporate as much knowledge into the design as possible. This knowledge may relate to the target protein itself or may be derived from known active and inactive compounds. Other factors should also be taken into account, such as the cost of the library and the physicochemical properties of the compounds that are contained within the library. Thus, library design is a multi objective optimization problem. Most approaches to optimizing multiple objectives are based on aggregation methods whereby the objectives are assigned relative weights and are combined into a single fitness function. A more recent approach involves the use of a Multiobjective Genetic Algorithm in which the individual objectives are handled independently without the need to assign weights. The result is a family of solutions each of which represents a different compromise in the objectives. Thus, the library designer is able to make an informed choice on an appropriate compromise solution.