QSAR Study On Hallucinogenic Phenylalkylamines

Hallucinogens are psychoactive substances that powerfully alter perception,mood,and a host of cognitive processes.Phenylalkylamines are a chemical class of hallucinogens.The study of their quantitative structureactivity relationship(QSAR) can be used to predict the hallucinogenic activity of analogs.It can also be used to classify and control abused drugs, and it is necessary to develop clinical therapeutic drugs.The study has practical guiding significance.In this paper,quantum chemical parameters of phenylalkylamines were calculated at the theoretical level of DFT/B3LYP/ 6-311+G(d,p) and two topology parameters were computed through Chem3D program.Combing some empirical parameters,QSAR model of phenylalkylamines was constructed using artificial neural network(ANN) and multiple linear regression(MLR) methods and some results were rather satisfactory. The detailed contents are as follows:1 Adopting the experimental structure of phenylalkylamines as initial conformation,precise spacial configuration and eletronic structure parameters of 75 phenylalkylamines were calculated by Gaussian 03 program at the level of DFT/B3LYP/6-311+G(d,p).The obtained parameters are as follows:α(molecular polarizability),μ(molecular dipole moment),E_HOMO (energy of the highest occupied molecular orbit(HOMO)),E_LUMO(energy of the lowest unoccupied molecular orbit(LUMO)),â–³E_HL(the difference in energy between the HOMO and LUMO),E_SHOMO(energy of the next highest occupied molecular orbit(SHOMO)),E_SLUMO(energy of the next lowest unoccupied molecular orbit(SLUMO)),â–³E_H(the difference in energy between the HOMO and SHOMO),â–³E_L(the difference in energy between the SLUMO and LUMO),and Q_o(sum of net charges on the ortho carbon atoms), Q_m(sum of net charges on the meta carbon atoms),Q_p(net charge on the para carbon atom),Q_r(total charge on the benzene ring carbon atoms),and Q_n(net charge on the nitrogen atom). In addition,two distance-based topological indices——Wiener(W) and Balaban(J) indices of phenylalkylamines were calculated with Chem Office8.0 Software,and sum of the hydrophobicities of the substituents on the ortho positions(H_o),sum of the hydrophobicities of the substituents on the meta positions(H_m),hydrophobicity of the substituent on the para position(H_p) were obtained through manual calculation.Total hydrophobicity (H) and hydrophobicities of substitutents during calculating were the values quoted by literature.Considering presence or absence ofα-methyl group,indicator variable I_me was uesd,which takes the value 1 if there isα-methyl group on theαcarbon atom,and 0 otherwise.2 The quantum chemical study of 75 phenylalkylamines was carried out at two aspects of frontier molecular orbitals and charge distributions.It could be inferred when phenylalkylamines interact with the acceptor they act as electron donors.The substituted benzene ring and the substituent sulfur or oxygen atom on the ortho positions,meta positions,or para position are active sites of phenylalkylamines.3 33 compounds which are similar in the structure were selected from 75 phenylalkylamines and were analyzed with multiple linear regression.6 variables which obviously influence hallucinogenic activity of phenylalkylamines were selected by backward eliminating variables.They areα,Q_n,Q_r, H_o,H_m,and J.Subsequently,linear regression equation was set up as follow:logMU=0.468α-7.627Q_n -2.104Q_r -5.157H_o -0.510H_m -0.0000063J-12.488 For this equation,the correlation coefficient R is 0.9340 and the standard error Se is 0.2068.The linear equation was tested with cross-validation through the leave-one-out(LOO) procedure and the complex correlation coefficient was obtained(R_CV² = 0.7920).It proves favorable stability and prediction performance of the linear model.9 compounds were selected stochastically from 75 phenylalkylamines and their activities were predicted by above equation.The predicted effect to the external samples was acceptable.The linear equation shows that the benzene ring and the nitrogen atom should be the active sites of phenylalkylamines and they would interact with the acceptor acting as electron donors.The results are basically consistent with that of quantum chemistry.4 For 33 phenylalkylamines,6 parameters(α,Q_n,O_r,H_o,H_m,J) were used as input and the QSAR model was set up with three-layer BP neural network(6-10-1).The correlation coefficient of the model was calculated to be R=0.9992 and the standard error Se=0.0036.The model was tested with cross-validation through the leave-one-out(LOO) procedure and the complex correlation coefficient was R_CV² =0.7250.That indicates that the fitting performance of ANN method is better than that of MLR.Although the predicted effect to the internal samples is inferior to MLR model,it could be acceptable.9 compounds as external predicted group were selected stochastically from 75 phenylalkylamines and their activity were predicted by the ANN model.It could be found that the predicted effect to the external samples by the ANN model was comparatively well also.So the artificial neural network could be acted as an effective QSAR modeling tool to explore the hidden nonlinear relationship between the biological activities and structural parameters.5 For 75 phenylalkylamines,because the result of MLR method was rather too poorish,artificial neural network method was used for the QSAR modelling study.First,7 parameters(Q_o,H_p,Q_p,I_me,â–³E_H,μ,H_o) which are the most influential factors on hallucinogenic activity of 75 phenylalkylamines were selected by stepwise regression analysis method.Then they were used as the input and the ANN(6-10-1) QSAR model was set up.Its correlation coefficient is R=0.9752 and the standard error is Se=0.0128.The model was tested with cross-validation through the LOO procedure and the complex correlation coefficient was R_CV² = 0.6972.The results show that the fitting effect of the model is considerable well,and stability and prediction performance of the model are favorable.This indicates that for 75 phenylalkylamines influencing factors on hallucinogenic activity could be nonlinear.6 Based on the linear regression results in literature,49 phenylalkylamines and 36 2-phenylindole derivatives were studied with artificial neural network method.And the results of ANN models were compared with that of MLR models.It could be concluded that the fitting effect and prediction performance of the ANN models were better.