Application of FLIP Technology

The parameters that were varied in order to study their effects on data mining of the active molecules included (1) number of conformations of virtual com- 

The Ochiai similarity coefficient with 250 maximum conformations provided the best retrieval rates of the actives from the virtual library (Fig. 9.5). 

The non-binary all feature combination of the 3D fingerprint and interaction feature was able to retrieve -90% of the actives on screening 20% of the database. The Ochiai coefficient was slightly better than Tanimoto and significantly better than Dice and Hamming in retrieving actives. All features for the enzyme 

In this exercise for the FGF target on the seeded database, we retrieved 70-80% of actives by screening 20-30% of the database. The best combination for retrieving maximum actives when 5% of the database was screened was nonbinary-all feature-Ochiai similarity coefficient . The effect of conformations is dependent on the flexibility of the molecules in the database. Based on our analysis, the maximum number of conformations set to 100 was sufficient for retrieving 70-80% of the actives. The hit list is 12-16 times enriched with respect to random selection from the database on screening only 5% of the database. 

Preliminary data on the selectivity issue are shown in Fig. 9.6. The virtual library comprised 1000 molecules from ACD (ACD v. 2000-1, Molecular Design) as described previously. Ten actives for FGF and CDK2 proteins were added to the random molecules to generate the virtual library for screening. The goal of the experiment was to isolate selectively FGF actives from the virtual library 

---