QESAR study tripeptide analogues as antioxidation agents

A database consisting of 23 tripeptides was used to study the quantitative relationships between electric surface potential descriptors and antioxidant activity QESARs. The important structural descriptors SaaNH_acnt, SsOH_acnt, SaaN, SaaN_acnt, SsssCH, SaaaC, SsNH3p, SdO, SdO_acnt were selected for constructing the linear models QESARs with genetic algorithm. The best 4-variable linear model QESARlinear including the structural descriptors SaaN, SdO, SdO_acnt and SsOH_acnt was constructed. The quality QESARlinear was exhibited in statistical values R2 fitness of 97.5660, standard error of estimation SE of 0.0378, F-stat of 130.2731, R2 test of 93.3851. The non-linear model as neural network model QESARneural I(4)-HL(3)-O(1) with R2 fitness of 98.2296 was built by using structural descriptors in QESARlinear model. The antioxidation activities of tripeptides resulting from QESARlinear and QESARneural model were pointed out in values MARE, % of 27.4282 and 20.0672, respectively.

Tạp chí Đại học Thủ Dầu Một, số 3(5) – 2012 QESAR STUDY OF TRIPEPTIDE ANALOGUES AS ANTIOXIDATION AGENTS Nong Thi Hong Duyen(1) – Pham Van Tat(2) (1) Hue University of Science; (2) Thu Dau Mot University ABSTRACT A database consisting of 23 tripeptides was used to study the quantitative relationships between electric surface potential descriptors and antioxidant activity QESARs The important structural descriptors SaaNH_acnt, SsOH_acnt, SaaN, SaaN_acnt, SsssCH, SaaaC, SsNH3p, SdO, SdO_acnt were selected for constructing the linear models QESARs with genetic algorithm The best 4-variable linear model QESARlinear including the structural descriptors SaaN, SdO, SdO_acnt and SsOH_acnt was constructed The quality QESARlinear was exhibited in statistical values R2fitness of 97.5660, standard error of estimation SE of 0.0378, F-stat of 130.2731, R2test of 93.3851 The non-linear model as neural network model QESARneural I(4)-HL(3)-O(1) with R2fitness of 98.2296 was built by using structural descriptors in QESARlinear model The antioxidation activities of tripeptides resulting from QESARlinear and QESARneural model were pointed out in values MARE, % of 27.4282 and 20.0672, respectively Keywords: QESARs model, multiple regression, neural network and antioxidation tripeptides * Introduction xidation activities QESAR may indicate quantitatively change of biological activity The antioxidation compounds prevent or the biological and chemical substances from physicochemical properties corres- ponding to composition of amino acids in radical-induced oxidation damage [4] The peptide chain [2], [3] hydrolysis from various proteins, such as This soybean, casein, bullfrog, royal jelly, venison, work reports the use of r-lactalbumin, myofibrillar, rice endosperm, multivariate regression and neuro-fuzzy have been shown to have antioxidant technique activities against the peroxidation of lipids construct the quantitative relationships or radical scavenging activities [1] between with genetic electric algorithm surface to potential Relationships between structural desc- descriptors and antioxidation activities for riptors (electric surface potential) and antio- tripeptides The electric surface potential 11 Journal of Thu Dau Mot university, No3(5) – 2012 descriptors of tripeptides are calculated by adjusting the control 1.0) were taken from incorporating molecular mechanics MM+ a and chemical experimental data were divided into the calculation SCF PM3 The linear model training set as calibration group and the QESARlinear and model test set as external validation set The QESARneural are those validation set of tripeptides was derived semiempirical structural quantum non-linear founded descriptors The by source randomly antioxidant of Li Yao from Wang original [1] data The The remaining tripeptides were constituted the activities of tripeptides resulting from training these models QESARs are compared to set This set includes 18 tripeptides with values of experimental those from literature activities, as listed in Table The ACexp Methodology values in range 0.0441 – 0.6369 were used 2.1 Antioxidant data to fit for the adjustable parameters of 23 QESAR models The test set consisting of antioxidation tripeptides used in this study tripeptides in Table with ACexp values (ACexp: antioxidant activities of peptides in range 0.3170 – 0.6369 was used to were measured by the ferric thiocyanate evaluate its predictability The experimental data of methods which are relative activities by Table The tripeptide structures and experimental antioxidant values ACexp , respectively [1] No Tripeptide ACexp No Tripeptide ACexp CYY 0.4699 13 HHR 0.0635 HHA 0.0680 14 HHS 0.0862 HHC 0.1277 15 HHT 0.0862 HHD 0.1877 16 HKH 0.0441 HHE 0.1877 17 HRH 0.0441 HHG 0.3170 18 LWL 0.6061 HHI 0.0680 19 PWK 0.4066 HHK 0.0635 20 RWK 0.6061 HHL 0.0680 21 RWQ 0.6061 10 HHM 0.0817 22 RWV 0.6061 11 HHN 0.3170 23 YYC 0.6369 12 HHQ 0.3170 12 Tạp chí Đại học Thủ Dầu Một, số 3(5) – 2012 2.2 Electric surface potential descriptors 2.4 Neural networks The tripeptide structures were built Neural networks NNs are artificial and optimized by using MM+ molecular intelligent systems They use a large mechanics number of interrelated data-processing method and semi-empirical package neurons to emulate the function of brain HyperChem [5] The optimization was Although there are several NN models in performed by Polak-Ribiere algorithm at use today, the most frequently used type gradient level 0.05 Tripeptide notation I(i)-HL(m)-O(n) in this research consists of PM3 and calculation their activities level in experimental are presented three-layered back-propagation neural net antioxidant in Table In this neural net, the neurons are arranged in an input layer I(i) with i Program QSARIS [7] was used to calculate neurons, a hidden layer HL(m) with m the electric surface potential descriptors of neurons, and an output layer O(n) with n each tripeptide, respectively The electric surface potential descriptors neurons Each neuron in any layer is fully with connected with the neurons of another calculation techniques were pointed out in layer The neural net was trained by using literature [9] the 2.3 Regression analysis Results and discussion that MLR models can be obtained using a step-wise multiple regression procedure; 3.1 among these models, the best one must be parameters: descriptors, the the number of square surface molecular descriptors SE Multiple and linear number are very often estimating the regression descriptors and descriptors SaaN, SdO, SdO_acnt and SsOH_acnt were identified and included in of the QESARlinear model, and there was no regression used potential through linear regression analysis Four significant (MLR) techniques based on least-squares procedures linear constructed based on the training set model is one that has high R2 and F low and experimental antioxidant values was first correlation and the F-stat value A reliable MLR and selection The correlation between the electric statistical coefficient (R2), the standard Error (SE) values, Variable relationship chosen [8], [9] For this objective, it is four transfer 0.7 and random seed 10,000 [6] selection or model development It is clear consider sigmoid hidden layer, momentum 0.7, learning rate MLR procedure was used for variable to as function was applied to each node in the A step-wise multiple linear regression common parameters correlation between the selected descriptors for The coefficients electric surface potential using program packages Regress [8] and descriptors were selected by using the QSARIS [7], [9] linear regression techniques forward and 13 Journal of Thu Dau Mot university, No3(5) – 2012 back elimination The best-suitable model 0.0378 and F-stat of 130.2731 The t-Stat QESARlinear (1) with four variables was ratio values of coefficients in linear model selected QESARlinear to describe accurately the were tested by statistical quantitative relationship between electric criteria at confident level a = 0.05 These surface turn potential descriptors (X) and antioxidant values (Y) out to be very satisfactory for statistical standards This linear model QESARlinear (1) needs also to be validated AC = 0.4002 – 0.0753SaaN – 0.0671SdO + by 0.8702SdO_acnt – 0.0765SsOH_acnt (1) cross-validation and external validation The cross-validation results The linear model QESARlinear (1) with showed that linear model QESARlinear (1) k = was adopted with statistical value can be used to predict the antioxidant R2test of 93.3851 The quality of this model values of any tripeptides QESARlinear was also reflected by value 50 45 40 35 30 25 20 15 10 0.80 R2 = 0.975 0.70 ACexp Values MPxk , % R2fitness of 97.5660, standard error SE of of 0.60 0.50 0.40 0.30 0.20 0.10 0.00 SsOH_acnt SaaN SdO SdO_acnt 0.1 0.2 0.3 0.4 0.5 0.6 0.7 AC pred Predictors a) b) Figure a) Mean values of contribution percentage MPxk,%; b) Correlation of values ACexp versus ACpred of training set (o) and the test set (●) for QESARlinear model and QESARneural model (∆) SaaN > SdO The values Pmxk,% and Moreover the important contribution of model MPxk,% for each predictor in model (1) was QESARlinear (1) was arranged in order exhibited in Figure So, the important SdO_acnt > SdO > SaaN > SsOH_acnt contribution of each descriptor in this model These QESARlinear (1) molecular descriptors based on in the this mean values of contribution percentage MPxk,% [9] In this to each descriptor not on the The values Pmxk,% and MPxk,% in was Figure were calculated by following arranged in order SdO_acnt > SsOH_acnt > formula [9] Pm xk ,%  100 bm ,i xm,i C total  rely magnitude of the coefficient to make case the magnitude of regression coefficients orresponding may (2) N MPm xk ,%   100 bm ,i xm ,i C total N j 1  14 k with Ctotal = b i 1 m,k xm , k (3) Tạp chí Đại học Thủ Dầu Một, số 3(5) – 2012 Where N of layers I(4)-HL(3)-O(1) The input layer tripeptides in training set; and m of is I(4) involves four neurons SaaN, SdO, number SdO_acnt and SsOH_acnt The output of of 18 predictors is in number this model QESARlinear layer O(1) is only neuron ACexp The hidden 3.2 Neural network model layer HL(3) includes three neurons The quality of this non-linear The NN models were generated by model QESARneural appeared by value using four descriptors appearing in linear R2fitness of 98.2296 model QESARlinear (1) as their inputs One 3.3 Comparison of QESARlinear and neuron, which encoded the antioxidant QESARneural models activity, constituted the output layer, and the hidden layer contained a variable Predictability number of neurons QESARlinear and of linear model non-linear model QESARneural was validated carefully by The non-linear model as a NN model leave-one-out validation techniques The QESARneural was created by incorporating predicted the neuro-fuzzy technique with genetic antioxidation values of tripeptides in test set resulting from these algorithm in INForm system [[6]] This models, as shown in Table non-linear model type consists of three Table Experimental ACexp and predicted ACpred antioxidant activities of tripeptides No Tripeptide ACexp HHN linear model QESARlinear non-linear model QESARneural ACpred ARE, % ACpred ARE, % 0.3170 0.2491 21.4259 0.2856 9.9054 HHQ 0.3170 0.2255 28.8530 0.2570 18.9274 PWK 0.4066 0.6059 49.0205 0.5905 45.2287 RWQ 0.6061 0.7354 21.3278 0.5600 7.6060 YYC 0.6369 0.5317 16.5136 0.5180 18.6686 Value MARE, % 27.4282 20.0672 The predicted resulting from these The predicted values resulting from models was judged by absolute value of the these models QSARs were judged by relative error ARE, % [9], [10], the the absolute value of the relative error medium absolute value of the relative error ARE, %: MARE, % [9] was used for assessing ARE,%  100 (ACexp  AC pred )/ACexp (4) overall error of models QESAR 15 Journal of Thu Dau Mot university, No3(5) – 2012 Conclusion The medium absolute values of the relative error MARE, % were used for This work has appeared successfully assessing overall error for models QSARs: MARE,%  100 (AC exp  AC pred ) N AC exp the QESARlinear (5) test set; ACexp and ACpred descriptors are model non-linear model a set of molecular The turn non-linear out to be model better predictable than linear model QESARlinear out that the antioxidation values resulting The above results obtained from this work from linear model QESARlinear and non- can become a good research way and linear model QESARneural turn out to be promise not different (F = 0.0494 < F0.05 = 5.3177) has from QESAR QESARneural ANOVA one factor rating also pointed QESARneural model descriptors to establish the best-fitting values model and linear used to select consistently the important experimental and predicted antioxidant However, of QESARneural The Genetic algorithm was Where N of is number of tripeptides in construction for prediction of antioxidant activity values for tripeptides less MARE, % value than model QESARlinear * NGHIEÂN CỨU QESAR CỦA NHÓM TRIPEPTIDE NHƯ CÁC TÁC NHÂN CHỐNG OXI HÓA Nông Thò Hồng Duyên(1) – Phạm Văn Tất(2) (1) Trường Đại học Khoa học – Đại học Huế; (2) Trường Đại học Thủ Dầu Một TÓM TẮT Một sở liệu gồm 23 tripeptide sử dụng để nghiên cứu mối quan hệ đònh lượng tham số bề mặt tónh điện hoạt tính chống oxi hóa QESAR Các tham số cấu trúc quan troïng SaaNH_acnt, SsOH_acnt, SaaN, SaaN_acnt, SsssCH, SaaaC, SsNH3p, SdO, SdO_acnt chọn để xây dựng mô hình tuyến tính QESAR giải thuật di truyền Mô hình tuyến tính biến số tốt QESARlinear bao gồm tham số cấu trúc SaaN, SdO, SdO_acnt SsOH_acnt xây dựng Chất lượng mô hình QESARlinear thể giá trò thống kê R2fitness = 97,5660, sai số chuẩn ước tính SE = 0,0378, F-stat = 130,2731, R2test = 93,3851 Mô hình phi tuyến mô hình mạng rơron QESARneural cấu trúc I(4)-HL(3)-O(1) với R2fitness = 98,2296 xây dựng cách sử dụng tham số cấu trúc mô hình QESARlinear Các hoạt tính chống oxi hóa tripeptide nhận từ mô hình QESARlinear QESARneural cho thấy giá trò MARE, % = 27,4282 20,0672 tương ứng Từ khóa: mô hình QESAR, hồi qui bội, mạng thần kinh tripeptide chống oxi hóa 16 Tạp chí Đại học Thủ Dầu Một, số 3(5) – 2012 REFERENCES [1] Li Yao-Wang, Li B., He J., Qian P, J Molecular Structure, No 998, P 53–61, (2011) [2] S Mittermayr, M Olajos, T Chovan, G.K Bonn, A Guttman, Trends in Analytical Chemistry, Vol 27, No 5, (2008) [3] K Saito, J Dong-hao, T Ogawa, K Muramoto, E Hatakeyama, T Yasuhara, and K Nokihara, J Agric Food Chem., No.51, 3668#3674, (2003) [4] Zhang H Z., Yang D P and Tang G Y., Vol 11 (15/16), P 749 – 754 (2006) [5] HyperChem Release 8.05, Hypercube Inc., USA (2008) [6] INForm v2.0, Intelligensys Ltd., UK (2000) [7] QSARIS 1.1, Statistical Solutions Ltd., USA (2001) [8] D D Steppan, J Werner, P R Yeater, Essential Regression and Experimental Design for Chemists and Engineers, (2000) [9] Pham Van Tat, Development of Quantitative Structure-Activity Relationship and Quantitative Structure-Property Relationship, Natural science and technology publisher, Hanoi, (2009) [10] Pham Van Tat, Pham Thi Tra My, Vietnamese Journal of Chemistry and Application, P 10-15, No 4, (2010) 17 ... linear model training set as calibration group and the QESARlinear and model test set as external validation set The QESARneural are those validation set of tripeptides was derived semiempirical... of antioxidation tripeptides used in this study tripeptides in Table with ACexp values (ACexp: antioxidant activities of peptides in range 0.3170 – 0.6369 was used to were measured by the ferric... the QESARlinear model, and there was no regression used potential through linear regression analysis Four significant (MLR) techniques based on least-squares procedures linear constructed based