Checkpoint kinase 2 (Chk2) has a great effect on DNA-damage and plays an important role in response to DNA double-strand breaks and related lesions. activity and it is represented as a real number. We view the function as the score function. Since only assigns a number not a set of numbers in this work no rank function would be used for an inhibitor. Therefore the rank combination and the rule (b) in Remark 1 are not considered in the study. Suppose we have prediction results (scoring functions). There are combinatorially 2? Sodium Channel inhibitor 1 1 combinations for all individual prediction results with score functions. The total number of combinations to be considered for predicting biological activity of an inhibitor is 2? 1. This number of combinations can become huge when the number of prediction results is large. Moreover we have to evaluate the predictive power of each combination across all inhibitors. This study would start with combining only two prediction results which still retain fairly good prediction power. Suppose prediction results = 1 2 … = Best Fast Caesar that is BesttrainBesttest) generated for testing set inhibitors. Using data fusion results from various prediction results are combined to obtain predictions with larger accuracy rate. The diversity rank/score function is used to select the most suitable prediction results for combination. If these Trdn three best PhModels were selected there are nine prediction results and then there are 29 ? 1 = 511 combinations. According to the rule (a) (1) in Remark 1 the in the testing set = {and ? prediction results selected (in this study = 6) there are (in this study the number is 15) diversity score functions. If we let vary and fix the prediction result pair (= {is in = {1 2 3 … is different from the set which is the testing set considered. The set is used as the index set for the diversity rank function value and |is indeed the cardinality of inhibitors and is independent of the specific inhibitor under study. For two prediction results and ? 1)/2 diversity rank/score graphs to see which pair of prediction results would give the larger diversity measurement according to the rule (a) (2) in Remark 1. 2.5 Database Screen After examining 15 diversity rank/score graphs the PhModels and determined from the best prediction result pair were used to screen the NCI database for new Chk2 inhibitor candidates. Under the PhModel pharmacophore hypothesis screening can be used to screen small molecule database to retrieve the compounds as potential inhibitors that fit the pharmacophoric features. In this study the “Search 3D Database protocol” with the Best/Fast/Casear Search option in Accelrys Discovery Studio 2.1 Sodium Channel inhibitor 1 was employed to search the NCI database with 260 71 compounds. We could filter out and select the compounds in the NCI database based on the estimated activity and chemical features of PhModel. 2.6 Molecular Docking After the database screening approach the selected compounds can be further estimated according to the interaction energy between a receptor and a ligand through the molecular docking approach. In this study selected compounds in the NCI database were docked into Chk2 active sites by CDOCKER docking Sodium Channel inhibitor 1 program and then their CDOCKER interaction energies were estimated. Finally new potential candidates were retrieved from the NCI database with high interaction energy. The workflow of database screening and molecular docking approach was shown in Figure 4. Sodium Channel inhibitor 1 Figure 4 The workflow of database screening and molecular docking approach for new Chk2 inhibitor candidates. 3 Results 3.1 PhModel Generation Results Each of the ten PhModels using 25 training set inhibitors and HypoGen Best Fast and Caesar algorithms was generated by selecting hydrogen bond acceptor (A) hydrogen bond donor (D) and hydrophobic (H) and hydrophobic aromatic (HYAR) features. Each of the best PhModels Besttrain Fasttrain and Caseartrain was evaluated with the best rtrain and the predicted biological activities of training set inhibitors and rtrain were listed in Table 1 respectively. From Table 1 the Besttrain obtained better rtrain of value 0.955 than those by Fasttrain and Caseartrain. Moreover.