Advantages of Product-Based Optimization

Diverse sets of building blocks do not lead necessarily to an optimized diversity of the resulting library. The properties of the assembled final molecules are not a simple sum of the fragment properties. Especially physico-chemical constants (e.g., lipophi-licity, basicity, etc.) depend very much on interactions between structural fragments. It has been shown that assessing the diversity of the enumerated library leads to significantly more diverse libraries than from assessing merely the diversity of the reactants [129, 130]. 

Once the huge virtual library is enumerated and described, it delivers very valuable knowledge about the properties of the included compounds and about the accessible chemical space. 

The first screening library can be designed in terms of maximum diversity and high drug-likeness (drug-like educts are no guarantee for drug-like products ). 

If active compounds are identified in the library by high-throughput screening, focused libraries can be easily designed from the same virtual Hbrary. In this case the compounds are no longer selected by diversity maximization, but by looking for similarity to the hit compounds. 

Selection algorithms must be able to select building blocks in a way that an optimized set of products is generated by combinatorial explosion. Genetic or evolutionary algorithms are most commonly used for this optimization step [70, 101, 109, 130]. 

---