Article

Non-linear quantitative structure-activity relationship for adenine derivatives as competitive inhibitors of adenosine deaminase.

Department of Biophysics, Faculty of Science, Tarbiat Modares University, P.O. Box: 14115/175, Tehran, Iran.
Biochemical and Biophysical Research Communications (Impact Factor: 2.28). 01/2006; 338(2):1137-42. DOI: 10.1016/j.bbrc.2005.10.049
Source: PubMed

ABSTRACT Logistic regression and artificial neural networks have been developed as two non-linear models to establish quantitative structure-activity relationships between structural descriptors and biochemical activity of adenosine based competitive inhibitors, toward adenosine deaminase. The training set included 24 compounds with known k(i) values. The models were trained to solve two-class problems. Unlike the previous work in which multiple linear regression was used, the highest of positive charge on the molecules was recognized to be in close relation with their inhibition activity, while the electric charge on atom N1 of adenosine was found to be a poor descriptor. Consequently, the previously developed equation was improved and the newly formed one could predict the class of 91.66% of compounds correctly. Also optimized 2-3-1 and 3-4-1 neural networks could increase this rate to 95.83%.

0 Bookmarks
 · 
73 Views
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Combinations of multiple linear regressions, genetic algorithms and artificial neural networks were utilized to develop models for seeking quantitative structure-activity relationships that correlate structural descriptors and inhibition activity of adenosine deaminase competitive inhibitors. Many quantitative descriptors were generated to express the physicochemical properties of 70 compounds with optimized structures in aqueous solution. Multiple linear regressions were used to linearly select different subsets of descriptors and develop linear models for prediction of log(k(i)). The best subset then fed artificial neural networks to develop nonlinear predictors. A committee of six hybrid models - that included genetic algorithm routines together with neural networks - was also utilized to nonlinearly select most efficient subsets of descriptors in a cross-validation procedure for nonlinear log(k(i)) prediction. The best prediction model was found to be an 8-3-1 artificial neural network which was fed by the most frequently selected descriptors among these subsets. This prediction model resulted in train set root mean sum square error (RMSE) of 0.84 log(k(i)) and prediction set RMSE of 0.85 log(k(i)) (both equivalent of 0.10 in normal range of log(k(i))) and correlation coefficient (r(2)) of 0.91.
    FEBS Letters 03/2007; 581(3):506-14. · 3.34 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: In this study we investigated the nature of the active site on the enzyme adenosine deaminase (ADA) in hopes of finding an analog structure to act as an inhibitor. Our research involved comparing the binding patterns of adenosine to several structural analogs by analyzing and comparing kinetic and spectroscopic data. We examined variations in positions 2 and 6 of the carbon ring to determine if either structural change may result in inhibition. 2-chloroadenosine was used to examine the 2-position, and similarly, 6- chloroadenosine and N-6-cyclohexyladenosine were used to consider the 6-position. The Vmax and KM values were established from data analysis with the Michaelis-Menton equation, and were used to compare adenosine and 6-chloroadenosine, which is observed to act as a substrate. The KM for adenosine was found to be 11.85 µM and the Vmax was determined to be 6.8x10-9 M/s. Both 2-chloroadenosine and N-6-cyclohexyladenosine were found not to bind at all, suggesting that structural alterations at either location would not yield inhibition. However, the possibility exists of smaller alterations at the 6 position leading to ADA inhibition.
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Motivation: So far various statistical and machine learning tech-niques applied for prediction of β-turns. The majority of these tech-niques have been only focused on the prediction of β-turn location in proteins. We developed a hybrid approach for analysis and predic-tion of different types of β-turn. Results: A two-stage hybrid model developed to predict the β-turn types I, II, IV and VIII. Multinomial logistic regression was initially used for the first time to select sig-nificant parameters in prediction of β-turn types using a self-consistency test procedure. The extracted parameters were consisted of 80 amino acid positional occurrences and 20 amino acid percentages in β-turn sequence. The most sig-nificant parameters were then selected using multinomial logistic regression model. Among these, the occurrences of glutamine, histidine, glutamic acid and arginine, respec-tively, in positions i, i+1, i+2 and i+3 of β-turn sequence had an overall relationship with 5 β-turn types. A neural network model was then constructed and fed by the parameters se-lected by multinomial logistic regression to build a hybrid predictor. The networks have been trained and tested on a non-homologous dataset of 565 protein chains by nine fold cross-validation. It has been observed that the hybrid model gives a Matthews correlation coefficient (MCC) of 0.235, 0.473, 0.103 and 0.124, respectively, for β-turn types I, II, IV and VIII which are best among previously reported re-sults. Our model also distinguished the different types of β-turn in the embedded binary logit comparisons which have not carried out so far.