Genetic Function Approximation Model Research Articles

Minimum Ignition Energy (MIE) prediction models for ignition sensitive fuels using machine learning methods

Quantitative Structure-Property Relationship (QSPR) regression models for Minimum Ignition Energy (MIE) prediction of 60 flammable compounds have been developed using machine learning algorithms Random Forests (RF), Artificial Neural Networks (ANN) and a non-machine learning algorithm - Genetic Function Approximation (GFA).RF algorithm was implemented for feature selection to identify the 13 molecular descriptors having maximum effect on MIE prediction accuracy (i.e. descriptors affecting MIE prediction accuracy >1%). Thereafter, these descriptors were used as input parameters to develop the RF, ANN and GFA models. The optimized RF algorithm resulted in test set R2 of 0.85 and displayed high internal robustness and external predictability. The ANN and GFA algorithms displayed improved performance only on addition of an additional parameter - Structure Parameter to the existing 13 descriptor set. The optimized ANN and the GFA models displayed a test set R2 of 0.79 and 0.71, respectively. The ANN model resulted in lower machine learning bias as compared to the RF model. The ANN model was observed to be more reliable than RF model for MIE prediction of small datasets. Based on this work, the RF algorithm for feature selection in QSPR modeling and the ANN, RF algorithms were observed to be promising options for MIE prediction applications.

QSAR modeling and in silico design of small-molecule inhibitors targeting the interaction between E3 ligase VHL and HIF-1α.

Protein-protein interactions (PPIs) have attracted much attention recently because of their preponderant role in most biological processes. The prevention of the interaction between E3 ligase VHL and HIF-1[Formula: see text] may improve tolerance to hypoxia and ameliorate the prognosis of many diseases. To obtain novel potent inhibitors of VHL/HIF-1[Formula: see text] interaction, a series of hydroxyproline-based inhibitors were investigated for structural optimization using a combination of QSAR modeling and molecular docking. Here, 2D- and 3D-QSAR models were developed by genetic function approximation (GFA) and comparative molecular field analysis (CoMFA) and comparative molecular similarity index analysis (CoMSIA) methods, respectively. The top-ranked models with strict validation revealed satisfactory statistical parameters (CoMFA with [Formula: see text], 0.637; [Formula: see text], 0.955; [Formula: see text], 0.944; CoMSIA with [Formula: see text], 0.649; [Formula: see text], 0.954; [Formula: see text], 0.911; GFA with [Formula: see text], 0.721; [Formula: see text], 0.801; [Formula: see text], 0.861). The selected five 2D-QSAR descriptors were in good accordance with the 3D-QSAR results, and contour maps gave the visualization of feature requirements for inhibitory activity. A new diverse molecular database was created by molecular fragment replacement and BREED techniques for subsequent virtual screening. Eventually, 31 novel hydroxyproline derivatives stood out as potential VHL/HIF-1[Formula: see text] inhibitors with favorable predictions by the CoMFA, CoMSIA and GFA models. The reliability of this protocol suggests that it could also be applied to the exploration of lead optimization of other PPI targets.

QSAR model for prediction of the therapeutic potency of N-benzylpiperidine derivatives as AChE inhibitors

A new family of AChE inhibitors, N-benzylpiperidines, showed exceptional efficacy in vitro and in vivo, minimal side effects and high selectivity for acetylcholinesterase (AChE). Three regression methods were chosen in this work to develop robust predictive models, namely multiple linear regression (MLR), genetic function approximation (GFA) and multilayer perceptron network (MLP). Ten descriptors were selected for a dataset of 99 molecules, using a genetic algorithm. The best results were obtained for MLP with a 10-6-1 artificial neural network model trained with the Broyden–Fletcher–Goldfarb–Shanno (BFGS) algorithm. Statistical prediction for MLR and GFA were r2 = 0.882 and r2 = 0.875, respectively. Because internal and external validation strategies play an important role, we adopted all available validation strategies to check the robustness of the models. All criteria used to validate these models revealed the superiority of the GFA model. Therefore, the models developed in this study provide an excellent prediction of the inhibitory concentration of a new family of AChE inhibitors.

Computational study of the effects of cations and anions to the cytotoxicity of diverse ionic liquids by supervised machine learning

Ionic liquids (ILs) have been widely used in many fields due to their unique physicochemical properties, and even they were considered as green solvents. However, the recent researches showed that ILs might bring potential risk to environment and humans. In this work, genetic function approximation (GFA) and least squares support vector machine (LSSVM) models were developed for predicting the cytotoxicity of a great variety of ILs, including 9 types of cations and 44 types of anions, to Leukemia Rat Cell Line (IPC-81) based on the structural descriptors calculated from the combination of cations and anions. Seven descriptors were selected by GFA to develop the linear QSAR model. According to the discussion of descriptors, the cation structure was the main factor of the toxicity which mainly depends on the hydrophobicity and space structure of cations. The LSSVM model was built to predict accurately the cytotoxicity of ILs to capture the nonlinear nature. The rigorous internal and external validation and applicability domain (AD) were performed to verify the reliability and predictability for GFA and LSSVM models. The results indicated that both models could be used for estimating the cytotoxicity of new ILs to IPC-81, and the discovered key structural characteristics could provide reference information for designing and synthesizing safer ILs.

Molecular dynamics simulated validation of anti-cancerous alkaloids as Topo IIβ inhibitors screened by QSAR, pharmacophore and molecular docking approaches

Looking to the recent roles of topoisomerases in various types of cancers, in the present paper, a quantitative structure–activity relationship model, based on genetic function approximation (GFA) method, was developed using 41 olivacine derivatives as inhibitors of topoisomerase IIβ (Topo IIβ). The best predictive GFA model explained the biological activity of the training and test sets with correlation coefficient values (r 2) of 0.747 and 0.549, respectively, and a significant cross-validated correlation coefficient (q 2) of 0.525. The model suggested positive correlation between activity and descriptors, namely partition coefficient (ALogP), electrostatic energy, number of hydrogen bond acceptors and number of rotatable bonds, while those of negative correlations with Jurs RPCG, Jurs TASA and PHI. Furthermore, a common feature-based pharmacophore model for these inhibitors was also developed which comprised of five features, namely one H-bond acceptor, two ring aromatic, one positive ionizable and one hydrophobic group. Screening of a large database for natural compounds, using both these developed models, led to identification of the pyrrole derivative, namely 1-(2-(dimethylamino)ethyl)-3-hydroxy-4-(4-methoxy-2-methylbenzoyl)-5-(3,4,5-trimethoxyphenyl)-1H-pyrrol-2(5H)-one (compound ID: STOCK1N-31995) with a predicted IC50 value of 0.3 × 10−6 µM as most potent inhibitor of Topo IIβ. The result of virtual screening was further validated using molecular dynamics (MD) simulation analysis. Thus, a 15 ns MD simulation analysis revealed high stability and effective binding of 1-(2-(dimethylamino)ethyl)-3-hydroxy-4-(4-methoxy-2-methylbenzoyl)-5-(3,4,5-trimethoxyphenyl)-1H-pyrrol-2(5H)-one within the active site of Topo IIβ.

Predicting the ecotoxicity of ionic liquids towards Vibrio fischeri using genetic function approximation and least squares support vector machine

Ionic liquids (ILs) are widely used in industrial production for their unique physicochemical properties, and they are even regarded as green solvents. However, the recent study showed ILs might pose a potential risk to aquatic ecosystems. In the present work, the quantitative structure–activity relationship (QSAR) models, including genetic function approximation (GFA) and least squares support vector machine (LSSVM) were developed for predicting the ecotoxicity of ILs towards the marine bacterium Vibrio fischeri based on the descriptors calculated from cations and anions. Five descriptors were selected by GFA and used to develop the linear model. From the discussion of descriptors, the cation structure was the main factor to the toxicity, which mainly depended on the size, lipophilic, and 3D molecular structure of cations. In order to capture the nonlinear nature, the LSSVM model was also built for more accurately predicting the ecotoxicity. The GFA and LSSVM models were performed the rigorous internal and external validation, further verifying these models with excellent robustness and predictive ability. Therefore, both of models can be used for the prediction of the ecotoxicity of newly synthesized and untested ILs, and can provide reference information and theoretical guidance for designing and synthesizing safer and more eco-friendly ILs.

Prediction of Density, Surface Tension, and Viscosity of Quaternary Ammonium-Based Ionic Liquids ([N222(n)]Tf2N) by Means of Artificial Intelligence Techniques

In this work, thermophysical properties of quaternary ammonium-based ionic liquids (ILs) including density, surface tension, and viscosity are produced by two powerful artificial intelligence techniques: genetic function approximation (GFA) and artificial neural network (ANN). In proposed GFA and ANN models, the critical temperature and water content of studied ILs ([N222(n)]Tf2N with n = 5, 6, 8, 10, and 12) as well as operation temperature were given as the input parameters and the density, surface tension, and viscosity were predicted as the output results. The obtained results reveal that the selected input parameters are appropriate for prediction of thermophysical properties of quaternary ammonium-based ILs. In addition, the high statistical quality represented by various criteria and the low prediction errors of the presented models indicate that they can accurately predict the density, surface tension, and viscosity of new ILs without recourse to experimental data.

Combination of pharmacophore hypothesis, genetic function approximation model, and molecular docking to identify novel inhibitors of S6K1

S6K1 has emerged as a potential target for the treatment for obesity, type II diabetes and cancer diseases. Discovery of S6K1 inhibitors has thus attracted much attention in recent years. In this investigation, a hybrid virtual screening method that involves pharmacophore hypothesis, genetic function approximation (GFA) model, and molecular docking technology has been used to discover S6K1 inhibitors especially with novel scaffolds. The common feature pharmacophore hypothesis and GFA regression model of S6K1 inhibitors were first developed and applied in a virtual screen of the Specs database for retrieving S6K1 inhibitors. Then, the molecular docking method was carried out to re-filter these screened compounds. Finally, 60 compounds with promising S6K1 inhibitory activity were carefully selected and have been handed over to the other group to complete the follow-up compound synthesis (or purchase) and activity test.Electronic supplementary materialThe online version of this article (doi:10.1007/s11030-013-9473-7) contains supplementary material, which is available to authorized users.

Computational Studies of Novel Chymase Inhibitors Against Cardiovascular and Allergic Diseases: Mechanism and Inhibition

To provide a new idea for drug design, a computational investigation is performed on chymase and its novel 1,4-diazepane-2,5-diones inhibitors that explores the crucial molecular features contributing to binding specificity. Molecular docking studies of inhibitors within the active site of chymase were carried out to rationalize the inhibitory properties of these compounds and understand their inhibition mechanism. The density functional theory method was used to optimize molecular structures with the subsequent analysis of highest occupied molecular orbital, lowest unoccupied molecular orbital, and molecular electrostatic potential maps, which revealed that negative potentials near 1,4-diazepane-2,5-diones ring are essential for effective binding of inhibitors at active site of enzyme. The Bayesian model with receiver operating curve statistic of 0.82 also identified arylsulfonyl and aminocarbonyl as the molecular features favoring and not favoring inhibition of chymase, respectively. Moreover, genetic function approximation was applied to construct 3D quantitative structure-activity relationships models. Two models (genetic function approximation model 1 r(2) = 0.812 and genetic function approximation model 2 r(2) = 0.783) performed better in terms of correlation coefficients and cross-validation analysis. In general, this study is used as example to illustrate how combinational use of 2D/3D quantitative structure-activity relationships modeling techniques, molecular docking, frontier molecular orbital density fields (highest occupied molecular orbital and lowest unoccupied molecular orbital), and molecular electrostatic potential analysis may be useful to gain an insight into the binding mechanism between enzyme and its inhibitors.

A QSAR Study of Biphenyl Analogues of 2-Nitroimidazo-[2, 1-b] [1, 3] - oxazines as Antitubercular Agents Using Genetic Function Approximation

A QSAR study was performed on ninety eight substituted biphenyl analogues of 2-nitroimidazo-[2, 1-b] [1, 3]-oxazines as antitubercular agents to explore the importance of topological, thermodynamic, spatial and physicochemical properties of the molecules towards the antitubercular activity. Genetic function approximation (GFA) was used as the chemometric tool for the study. The study shows that ortho and meta linked attachments of the biphenyl analogs to 2- nitroimidazo-[2, 1-b] [1, 3]-oxazines are detrimental for the antitubercular activity. Hydrophobicity, branching and presence of electronegative atoms enhance the activity. Based on the r(m)(2)((overall)) criterion, which considers both internal validation and external validation, a GFA model with spatial, thermodynamic and topological descriptors appears to be the best model (r(m)(2)((overall)) = 0.556).

Genetic Function Approximation and Bayesian Models for the Discovery of Future HDAC8 Inhibitors

Background: Histone deacetylase (HDAC) 8 is one of its family members catalyzes the removal of acetyl groups from N-terminal lysine residues of histone proteins thereby restricts transcription factors from being expressed. Inhibition of HDAC8 has become an emerging and effective anti-cancer therapy for various cancers. Application computational methodologies may result in identifying the key components that can be used in developing future potent HDAC8 inhibitors. Results: Facilitating the discovery of novel and potential chemical scaffolds as starting points in the future HDAC8 inhibitor design, quantitative structure-activity relationship models were generated with 30 training set compounds using genetic function approximation (GFA) and Bayesian algorithms. Six GFA models were selected based on the significant statistical parameters calculated during model development. A Bayesian model using fingerprints was developed with a receiver operating characteristic curve cross-validation value of 0.902. An external test set of 54 diverse compounds was used in validating the models. Conclusions: Finally two out of six models based on their predictive ability over the test set compounds were selected as final GFA models. The Bayesian model has displayed a high classifying ability with the same test set compounds and the positively and negatively contributing molecular fingerprints were also unveiled by the model. The effectively contributing physicochemical properties and molecular fingerprints from a set of known HDAC8 inhibitors were identified and can be used in designing future HDAC8 inhibitors.

Quantitative Structure–Property Relationship for Flash Points of Alcohols

In this paper, quantitative structure–property relationship (QSPR) models have been developed to predict flash points for some common alcohols based on a data set of 151 components. With the use of the genetic function approximation (GFA) approach, four descriptors have been selected from a set of more than 1000 molecular descriptors. These selected descriptors were used as inputs for the adaptive neuro-fuzzy inference system (ANFIS) model. The GFA model resulted in squared correlation coefficient values of 0.935 and 0.91 respectively for the training and test sets, whereas ANFIS resulted in the values of 0.959 and 0.951 for the training and test sets, respectively. However, the linear and nonlinear models can give satisfactory prediction results, but the ANFIS model is somewhat superior.

Quantitative structure–property relationship prediction of liquid thermal conductivity for some alcohols

Linear and nonlinear quantitative structure–property relationship (QSPR) models for prediction of liquid thermal conductivity of 116 alcohols were developed from a set of 1199 molecular descriptors by using genetic function approximation (GFA) and adaptive neuro-fuzzy inference system (ANFIS). Highly statistically significant model was obtained by GFA method when the number of descriptors in the equation was set to 5. Results of the GFA model were further compared with nonlinear QSAR model generated by ANFIS. The results surprisingly showed more or less the same quality for GFA and ANFIS modeling, according to the squared correlation coefficient values for testing set, which were 0.9521 and 0.9595, respectively.

Pharmacophore mapping, molecular docking and QSAR studies of structurally diverse compounds as CYP2B6 inhibitors

Pharmacophore mapping, molecular docking and quantitative structure–activity relationship (QSAR) studies were carried out for a structurally diverse set of 48 compounds as CYP2B6 inhibitors. The generated best pharmacophore hypotheses from the three methods of conformer generation (FAST, BEST and conformer algorithm based on energy screening and recursive buildup) indicate the importance of two features, namely, hydrogen bond acceptor [electron-rich centre] and ring aromaticity. The distance between the two centres of the important features for ideal inhibitors varied from 5.82 to 6.03 Å. The chemometric tools used for the QSAR analysis were genetic function approximation (GFA) and genetic partial least squares. The developed QSAR models indicate the importance of an electron-rich centre, size of molecule, impact of branching and ring system and distribution of charges in the molecular surface. The docking study confirms the importance of an electron-rich centre for binding with the iron atom of the cytochrome enzyme. A GFA model with spline option was found to be the best model based on internal validation as well as the r 2 m (overall) criterion (Q 2 = 0.772, r 2 m (overall) = 0.774). According to the external prediction statistics (R 2 pred = 0.876), another GFA-derived model with spline option outperforms the remaining models.

Exploring QSAR of hydroxyphenylureas as antioxidants using physicochemical and electrotopological state atom parameters

In the present study, free radical scavenging activity of 36 substituted hydroxyphenylurea derivatives was subjected to classical quantitative structure–activity relationship (QSAR) analyses using physicochemical (hydrophobicity and molar refractivity) and electrotopological state atom parameters. For the development of the QSAR models, statistical techniques such as stepwise multiple linear regression and genetic function approximation (GFA) were used. The developed models indicate an important contribution of the phenolic hydroxyl group of hydroxyphenylureas, apart from that of the urea moiety and piperazine nucleus of the side chain, to the free radical scavenging activity. The presence of substituents at the phenyl ring influences the electron density distribution over the phenolic ring system and modulates the activity. Hydrophobicity is found to contribute positively to the free radical scavenging activity. Based on internal validation (Q 2), external validation () and overall validation criteria (), a GFA model with spline options was found to be the best model (Q 2 = 0.957, = 0.966, = 0.914).

Docking and 3D-QSAR studies of diverse classes of human aromatase (CYP19) inhibitors

Aromatase (cytochrome 19) inhibitors have emerged as promising candidates for treatment of breast cancer. In search of potent aromatase inhibitors, docking and three-dimensional quantitative structure-activity relationship (3D-QSAR) studies using molecular shape, spatial, electronic, structural and thermodynamic descriptors have been performed on a diverse set of compounds having human aromatase inhibitory activities. An attempt has also been made to include two-dimensional (2D) descriptors in the QSAR studies. The chemometric tools used for model development are genetic function approximation (GFA) and genetic partial least squares (G/PLS). The docking study shows that the important interacting amino acids in the active site cavity are Met374, Arg115, Ile133, Ala306, Thr310, Asp309, Val370 and Ser478. One or more hydrogen bond formation with Met374 is one of the essential requirements for the ligands for optimum aromatase inhibition. The binding is further stabilized by van der Waals interactions with a few non-polar amino acid residues in the active site. The developed QSAR models indicate the importance of different shape, Jurs parameters, structural parameters, topological branching index and E-state index for different fragments. The results obtained from the QSAR analysis are supported by our docking observations. There should be one or two hydrogen bond acceptor groups (like -NO2, -CN) and optimal hydrophobicity for ideal aromatase inhibitors. A GFA model with spline option obtained using 3D descriptors was found to be the best model based on internal validation (Q2=0.668) while the best (externally) predictive model was a GFA model with spline option using combined set (2D and 3D) descriptors (Rpred2=0.687). Based on rm2(overall) criterion, the best model was a G/PLS model (using 3D descriptors) with spline option (rm2(overall)=0.606).

Exploring QSAR and QAAR for inhibitors of cytochrome P450 2A6 and 2A5 enzymes using GFA and G/PLS techniques

A series of naphthalene and non-naphthalene derivatives ( n = 42) having cytochrome P450 2A6 and 2A5 inhibitory activities, reported by Rahnasto et al., were subjected to QSAR and QAAR studies. The analyses were performed using electronic, spatial, shape and thermodynamic descriptors to develop quantitative models for prediction of the inhibitory activities and to explore importance of different descriptors for the responses. The data set was divided into training and test sets (with test set size being approximately 25% of the full data set size) based on K-means clustering applied on the standardized descriptor matrix. Genetic function approximation (GFA) and genetic partial least-squares (G/PLS) were used as chemometric tools for modeling, and the derived equations were of acceptable statistical and prediction (both internal and external) qualities although different equations varied in quality in a wide range ( R 2: 0.561–0.898, R a 2: 0.508–0.870, Q 2: 0.495–0.814, R pred 2: 0.615–0.914, r 2: 0.679–0.905, r 0 2: 0.639–0.904, r m 2: 0.494–0.876). In the case of CYP2A5 inhibition, the GFA derived QSAR model is better than the G/PLS derived model considering both internal and external validations. In the case of CYP2A6 inhibitory potency data, the GFA derived QSAR model is better than the G/PLS model considering internal validation whereas the latter is better in external validation (which is more important) than the former. The model development process was subjected to randomization test at 90% confidence level by taking into account the whole pool of descriptors, while the developed models were also subjected to randomization test (99% confidence level) for validation. Based on the randomization test results, GFA models are found to be superior to the G/PLS models. Among the parameters, which were found important in modeling both the responses, were different Jurs descriptors, electronic descriptors (like Sr, Apol), steric descriptors (like shadow indices, Molref), shape descriptors (like COSV, Fo) and lipophilicity descriptors. This indicates that the CYP2A5 and CYP2A6 inhibition of these compounds is related to charge distribution, surface area, electronic, hydrophobic and spatial properties of the molecules.