Table 2 CoMFA and CoMSIA results Model SE SEE bsr 92 SD CoMFA(standard) 0.947 0.30 94.645 0.265 0.019 CoMSIA(steric+electro) 0.705 0.861 0.512 71220 0.418 0.040 CoMSIA(steric+electro+ hydrophobic) 0.713 0.874 0.488 79.74 0.427 0.015 CoMSIA(steric+electro+ 0.682 0.465 0.335 0.938 CoMSIA(steric+electro+ Hydrophobictacceptor) 0.692 0.90 0.430 70.911 0.384 0.921 CoMSIA(all descriptors) 0.678 0.421 55.97 0.94 cr: noncross-validated correlation coefficient, dsEE standard error of estimate F: F-test value of estimate for boot strapping analysis, QSAR coefficient contour maps training set and the observed values versus predicted values in the test set. All pki values in the test set were The results of all the CoMFA and CoMSIA models were among the scope of the training set and both QSAr visualized using the ' Dev*Coeff mapping option. And models here were well validated from the test set methods olecule 17 was set as the template to validate these contour maps Characteristics of orvinol analogs binding to the K-opioid receptor Resuits and discussion From K-agonists that retain the full morphine structure, as CoMFA and CoMSIA models for k-opioid receptor well as the compounds in this study, some potent and elective k-agonists found through structural mod- In order to establish reliable 3D-QSAR models, analgesic ifications on two specific regions. Since compound 17([8] activities(in vitro ECso values) were also collected from in Fig. 1)was selected as the template for QSAr contour the literature. However, there was no obviously linear demonstration. The two special regions mapped to this relationship between pECso and pk; values(shown in compound were 6-substituted groups in the parent mor 3), and QSAR models established from pECso seemed phine structure and additional 19-substituted groups(see to be poor(q-=0. 269, N-2 and r=0.530 for 27 molecules Fig. 1). The influence of these substitutent groups on thei in CoMFA, data not shown), so only pki was used as the binding to the k-receptor are discussed below from QSAR dependent value. The reason was possibly that both contour plots CoMFA and CoMSIA methods are based on the assump- From the Electrostatic Fields contours shown in Fig 5a tion that changes in binding affinities of ligands are related (CoMFA contours), it is clear that any introduction of six to changes in molecular properties represented by multiple substitution or 6, 7-ring constrained structure could facilitate fields, steric, electrostatic, etc., and other factors besides binding only if they contain electronegative groups. This is binding affinity might be involved in determining the true since some potent K-agonists such as KT-95 [18] and analgesic potency of those compounds on the k-opioid TRK-820 [19](see Fig. 1)contain highly electronegative rec oxygen in this region. Interestingly, CoMSIA contours 3D-QSAR models of orvinol analogs were investigated showed replacing C-6 with more electron-donating groups by their binding affinities to the k-opioid receptor with K may favor binding to the k-receptor. However, no pre- values ranging from 0. 14 to 4841 nM. The best dominant electrostatic contours around the C-19 region were predictions were obtained by the CoMFA standard observed in Fig 5a model (=0.686, /=0.947)and CoMSIA combined Both yellow contours found in CoMFA and CoMSIA steric, electrostatic, hydrophobic, and donor/acceptor plots( Fig. 5b)showed that large steric groups may reduce fields((=0.678, /=0.914). All CoMFA and CoMSIA compound binding affinities in six substitutions.The analysis parameters and results are shown in Table 2. In binding affinities of all compounds with 6, 7-ring con- addition, Table 3 and Fig. 4 show the table and graph of strained indole structures decreased considerably when the observed values versus conventional fit values in the compared to the compounds without such structures in thiQSAR coefficient contour maps The results of all the CoMFA and CoMSIA models were visualized using the ‘stDev*Coeff’ mapping option. And molecule 17 was set as the template to validate these contour maps. Results and discussion CoMFA and CoMSIA models for κ-opioid receptor In order to establish reliable 3D–QSAR models, analgesic activities (in vitro EC50 values) were also collected from the literature. However, there was no obviously linear relationship between pEC50 and pKi values (shown in Fig. 3), and QSAR models established from pEC50 seemed to be poor (q 2 =0.269, N=2 and r 2 =0.530 for 27 molecules in CoMFA, data not shown), so only pKi was used as the dependent value. The reason was possibly that both CoMFA and CoMSIA methods are based on the assump￾tion that changes in binding affinities of ligands are related to changes in molecular properties represented by multiple fields, steric, electrostatic, etc., and other factors besides binding affinity might be involved in determining the analgesic potency of those compounds on the κ-opioid receptor. 3D–QSAR models of orvinol analogs were investigated by their binding affinities to the κ-opioid receptor with Ki values ranging from 0.14 to 4841 nM. The best predictions were obtained by the CoMFA standard model (q 2 =0.686, r 2 =0.947) and CoMSIA combined steric, electrostatic, hydrophobic, and donor/acceptor fields (q 2 =0.678, r 2 =0.914). All CoMFA and CoMSIA analysis parameters and results are shown in Table 2. In addition, Table 3 and Fig. 4 show the table and graph of the observed values versus conventional fit values in the training set and the observed values versus predicted values in the test set. All pKi values in the test set were among the scope of the training set and both QSAR models here were well validated from the test set methods. Characteristics of orvinol analogs binding to the κ-opioid receptor From κ-agonists that retain the full morphine structure, as well as the compounds in this study, some potent and selective κ-agonists were found through structural mod￾ifications on two specific regions. Since compound 17 ([8] in Fig. 1) was selected as the template for QSAR contour demonstration. The two special regions mapped to this compound were 6-substituted groups in the parent mor￾phine structure and additional 19-substituted groups (see Fig. 1). The influence of these substitutent groups on their binding to the κ-receptor are discussed below from QSAR contour plots: From the Electrostatic Fields contours shown in Fig. 5a (CoMFA contours), it is clear that any introduction of six substitution or 6, 7-ring constrained structure could facilitate binding only if they contain electronegative groups. This is true since some potent κ-agonists such as KT-95 [18] and TRK-820 [19] (see Fig. 1) contain highly electronegative oxygen in this region. Interestingly, CoMSIA contours showed replacing C-6 with more electron-donating groups may favor binding to the κ-receptor. However, no pre￾dominant electrostatic contours around the C-19 region were observed in Fig. 5a. Both yellow contours found in CoMFA and CoMSIA plots (Fig. 5b) showed that large steric groups may reduce compound binding affinities in six substitutions. The binding affinities of all compounds with 6, 7-ring con￾strained indole structures decreased considerably when compared to the compounds without such structures in this Table 2 CoMFA and CoMSIA results Model q2a N b r 2c SEEd F e SEE bsf q2 bsg SDh CoMFA(standard) 0.686 4 0.947 0.330 94.645 0.265 0.962 0.019 CoMSIA(steric+electro) 0.705 2 0.861 0.512 71.220 0.418 0.903 0.040 CoMSIA(steric+electro+ hydrophobic) 0.713 2 0.874 0.488 79.747 0.427 0.895 0.015 CoMSIA(steric+electro+ hydrophobic+donor) 0.682 3 0.895 0.465 62.345 0.335 0.938 0.032 CoMSIA(steric+electro+ Hydrophobic+acceptor) 0.692 3 0.906 0.430 70.911 0.384 0.921 0.030 CoMSIA(all descriptors) 0.678 4 0.914 0.421 55.975 0.337 0.946 0.019 a q2 : leave-one-out (LOO) cross-validated correlation coefficient, b N: optimum number of components, c r 2 : noncross-validated correlation coefficient, d SEE: standard error of estimate, e F: F-test value, f SEEbs: standard error of estimate for boot strapping analysis, g q2 bs: mean r squared of boot strapping analysis (ten runs), h SD: standard deviation 881