Investigations of the quantitative structure-activity relationships of a data set comprising 66 phenylalkanamines have been carried out using the CoMFA method. This yielded a cross-validated correlation coefficient (q'value) of more than 0.8. The target parameter used was the hallucinogenic effect on humans, since this variable is of particular importance for research into addictive substances. It was possible to confirm the reliability of the CoMFA analysis by using a second, independent phenylalkanamine data set. It was found that models with good predictive properties are obtained if up to ten components are taken into account. In a further step it was possible to include hallucinogenic tryptamine derivatives in a common QSAR analysis with the phenylalkanamines and this in spite of their differing basic structures. The final model from that the CoMFA plots were extracted is based on 148 compounds and permits precise inferences to be made concerning the relationships between structural elements and hallucinogenic effects.

A cannabinoid pseudoreceptor model for the CB1-receptor has been constructed for 31 cannabinoids using the molecular modelling software YAK. Additionally, two CoMFA studies were performed on these ligands, the first of which was conducted prior to the building of the pseudoreceptor. Its pharmacophore is identical with the initial superposition of ligands used for pseudoreceptor construction. In contrast, the ligand alignment for the second CoMFA study was taken directly from the final cannabinoid pseudoreceptor model. This altered alignment gives markedly improved cross-validated r2 values as compared to those obtained from the original alignment with r2 cross values of 0.79 and 0.63, respectively, for five components. However, the pharmacophore alignment has the better predictive ability. Both the CoMFA and pseudoreceptor methods predict the free energy of binding of test ligands well.

The 5-HT 2A receptor is known to act as the biological target for a series of hallucinogenic substances including substituted phenylalkylamines, tryptamines and LSD. A prerequisite for a hallucinogenic effect is an agonistic binding mode to the high-af®nity state of the receptor. Attempts to establish a quantitative structure-activity relationship for such compounds are typically based on homology models or 3D-QSAR. In this paper, we describe a surrogate for the 5-HT 2A receptor derived by means of quasi-atomistic receptor modeling (software Quasar), a more recently developed 3D-QSAR technique. This approach allows for the simulation of local induced ®t, H-bond ¯ip-¯op, and solvation phenomena. The QSARs are established based on a family of receptor-surface models, generated by a genetic algorithm combined with cross-validation. The surrogate for the 5-HT 2A receptor yielded a cross-validated q 2 of 0.954 for the 23 compounds de®ning the training set. A series of 7 test compounds was then used to validate the model, resulting in a RMS deviation of 0.40 kcalymol between DG 0 prd. and DG 0 exp.. The largest individual deviation was 0.61 kcaly mol, corresponding to an uncertainty of a factor 2.7 in the binding af®nity. A scramble test with negative outcome (q 2 ˆ 0.144, slope ˆ 7 0.019) demonstrates the sensitivity of the model with respect to the biological data. Subsequently, the surrogate was used to estimate the activity of a series of 53 hypothetical congeneric compounds, some of which are predicted to be close in activity to LSD.

The 5-HT 2A receptor is known to act as the biological target for a series of hallucinogenic substances including substituted phenylalkylamines, tryptamines and LSD. A prerequisite for a hallucinogenic effect is an agonistic binding mode to the high-af®nity state of the receptor. Attempts to establish a quantitative structure-activity relationship for such compounds are typically based on homology models or 3D-QSAR. In this paper, we describe a surrogate for the 5-HT 2A receptor derived by means of quasi-atomistic receptor modeling (software Quasar), a more recently developed 3D-QSAR technique. This approach allows for the simulation of local induced ®t, H-bond¯ip-¯op, and solvation phenomena. The QSARs are established based on a family of receptor-surface models, generated by a genetic algorithm combined with cross-validation. The surrogate for the 5-HT 2A receptor yielded a cross-validated q 2 of 0.954 for the 23 compounds de®ning the training set. A series of 7 test compounds was then used to validate the model, resulting in a RMS deviation of 0.40 kcalymol between DG 0 prd. and DG 0 exp.. The largest individual deviation was 0.61 kcaly mol, corresponding to an uncertainty of a factor 2.7 in the binding af®nity. A scramble test with negative outcome (q 2 0.144, slope 7 0.019) demonstrates the sensitivity of the model with respect to the biological data. Subsequently, the surrogate was used to estimate the activity of a series of 53 hypothetical congeneric compounds, some of which are predicted to be close in activity to LSD.