The advent of focused library and virtual screening has reduced the disadvantage of combinatorial chemistry and changed it to a realizable and cost-effective tool in drug discovery. Usually, genetic algorithms (GAs) are used to quickly finding high-scoring molecules by sampling a small subset of the total combinatorial space. Therefore, scoring functions play essential roles in focused library design. Reported here is our initial attempt to establish a new approach for generating a target-focused library using the combination of the scores of structural diversity and binding affinity with our newly improved drug-likeness scoring functions. Meanwhile, a software package, named LD1.0, was developed on the basis of the new approach. One test on a cyclooxygenase (COX)2-focused library successfully reproduced the structures that have been experimentally studied as COX2-selective inhibitors. Another test is on a peroxisome proliferator-activated receptors gamma-focused library design, which not only reproduces the key fragments in the approved (thiazolidinedione) TZD drugs, but also generates some new structures that are more active than the approved drugs or published ligands. Both of the two tests took approximately 15% of the running time of the ordinary molecular docking method. Thus, our new approach is an effective, reliable, and practical way for building up a properly sized focused library with a high hit rate, novel structure, and good ADME/T profile.