Comparing the Quality and Predictiveness between 3D QSAR Models Obtained from Manual and Automated Alignment

A comparison between 2D and 3D descriptors in QSAR modeling based on bio-active conformations

QSAR models are widely and successfully used in many research areas. The success of such models highly depends on molecular descriptors typically classified as 1D, 2D, 3D, or 4D. While 3D information is likely important, e. g., for modeling ligand-protein binding, previous comparisons between the performances of 2D and 3D descriptors were inconclusive. Yet in such comparisons the modeled ligands were not necessarily represented by their bioactive conformations. With this in mind, we mined the PDB for sets of protein-ligand complexes sharing the same protein for which uniform activity data were reported. The results, totaling 461 structures spread across six series were compiled into a carefully curated, first of its kind dataset in which each ligand is represented by its bioactive conformation. Next, each set was characterized by 2D, 3D and 2D + 3D descriptors and modeled using three machine learning algorithms, namely, k-Nearest Neighbors, Random Forest and Lasso Regression. Models' performances were evaluated on external test sets derived from the parent datasets either randomly or in a rational manner. We found that many more significant models were obtained when combining 2D and 3D descriptors. We attribute these improvements to the ability of 2D and 3D descriptors to code for different, yet complementary molecular properties.

AutoWeka: toward an automated data mining software for QSAR and QSPR studies

In biology and chemistry, a key goal is to discover novel compounds affording potent biological activity or chemical properties. This could be achieved through a chemical intuition-driven trial-and-error process or via data-driven predictive modeling. The latter is based on the concept of quantitative structure-activity/property relationship (QSAR/QSPR) when applied in modeling the biological activity and chemical properties, respectively, of compounds. Data mining is a powerful technology underlying QSAR/QSPR as it harnesses knowledge from large volumes of high-dimensional data via multivariate analysis. Although extremely useful, the technicalities of data mining may overwhelm potential users, especially those in the life sciences. Herein, we aim to lower the barriers to access and utilization of data mining software for QSAR/QSPR studies. AutoWeka is an automated data mining software tool that is powered by the widely used machine learning package Weka. The software provides a user-friendly graphical interface along with an automated parameter search capability. It employs two robust and popular machine learning methods: artificial neural networks and support vector machines. This chapter describes the practical usage of AutoWeka and relevant tools in the development of predictive QSAR/QSPR models.

AVAILABILITY

The software is freely available at http://www.mt.mahidol.ac.th/autoweka.

QSAR modeling of toxicity of acyclic quaternary ammonium compounds on Scenedesmus Quadricauda using 2D and 3D descriptors

Optimized calculation of typical acyclic quaternary ammonium compounds (QACs) was performed at B3LYP/6-311G** level using density functional theory (DFT) method. A two- dimensional quantitative structure-activity relationship (2D-QSAR) model was established with the obtained structure parameters as theoretical descriptors. And then three-dimensional quantitative structure-activity relationship (3D-QSAR) models were built using comparative molecular field analysis (CoMFA) and comparative molecular similarity indices analysis (CoMSIA) methods respectively. The 2D and 3D QSAR models exhibit optimum stability and predictive ability, revealing that steric and electronic effects influence the toxicity of acyclic QACs to Scenedesmus Quadricauda mostly.

X-ray crystallographic structures as a source of ligand alignment in 3D-QSAR

Three-dimensional quantitative structure-activity relationships (3D-QSAR) analyses are methods correlating a pharmacological property with a mathematical representation of a molecular property distribution around three-dimensional molecular models for a set of congeners. 3D-QSAR methods are known to be highly sensitive to ligand conformation and alignment method. The current study collects 32 unique positions of congeneric ligands co-crystallized with the binding domain of AMPA receptors and aligns them using protein coordinates. Thus, it allows for a unique opportunity to consider a ligands' orientation aligned by their mode of binding in a native molecular target. Comparative molecular field analysis (CoMFA) models were generated for this alignment and compared with the results of analogous modeling using standard structure-based alignment or obtained in docking simulations of the ligands' molecules. In comparison with classically derived models, the model based on X-ray crystallographic studies showed much better performance and statistical significance. Although the 3D-QSAR methods are mainly employed when crystallographic information is limited, the current study underscores the importance that the selection of inappropriate molecular conformations and alignment methods can lead to generation of erroneous models and false conclusions.

A combined 3D-QSAR and molecular docking strategy to understand the binding mechanism of (V600E)B-RAF inhibitors

B-RAF is a member of the RAF protein kinase family involved in the regulation of cell growth, differentiation, and proliferation. It forms a part of conserved apoptosis signals through the RAS-RAF-MAPK pathway. (V600E)B-RAF protein has much potential for scientific research as therapeutic target due to its involvement in human melanoma cancer. In this work, a molecular modeling study was carried out for the first time with 3D-QSAR studies by following the docking protocol on three different data sets of (V600E)B-RAF inhibitors. Based on the co-crystallized compound (PDB ID: 1UWJ), a receptor-guided alignment method was utilized to derive reliable CoMFA and CoMSIA models. The selected CoMFA model gives the best statistical values (q(2) = 0.753, r(2) = 0.962). With the same alignment protocol, a statistically reliable CoMSIA model out of fourteen different combinations was also derived (q(2) = 0.807, r(2) = 0.961). The actual predictive powers of both models were rigorously validated with an external test set, which gave satisfactory predictive r(2) values for CoMFA and CoMSIA models, 0.89 and 0.88, respectively. In addition, y-randomization test was also performed to validate our 3D-QSAR models. Contour maps from CoMFA and CoMSIA models supported statistical results, revealed important structural features responsible for biological activity within the active site and explained the correlation between biological activity and receptor-ligand interactions. Based on the developed models few new structures were designed. The newly predicted structure (IIIa) showed higher inhibitory potency (pIC(50) 6.826) than that of the most active compound of the series.

Acute toxicity and n-octanol/water partition coefficients of substituted thiophenols: determination and QSAR analysis

The acute toxicity (-log EC(50)) to Photobacterium phosphoreum and the n-octanol/water partition coefficient (log K(ow)) of 31 kinds of substituted thiophenols were determined at 298.15K. The -log EC(50) values of studied chemicals are between 4.26 and 5.89. Their log K(ow) values are between 1.34 and 4.02. Comparative molecular field (CoMFA) and comparative molecular similarity index analysis (CoMSIA) models established were successful in predicting -log EC(50) and log K(ow) values of halogenated, methylic, amino and methoxy thiophenols. The size of molecule is the main factor influencing the properties. No correlation was found between the properties and their structural and thermodynamic descriptors from DFT calculation.

Revealing interaction mode between HIV-1 protease and mannitol analog inhibitor

HIV protease is a key enzyme to play a key role in the HIV-1 replication cycle and control the maturation from HIV viruses to an infectious virion. HIV-1 protease has become an important target for anti-HIV-1 drug development. Here, we used molecular dynamics simulation to study the binding mode between mannitol derivatives and HIV-1 protease. The results suggest that the most active compound (M35) has more stable hydrogen bonds and stable native contacts than the less active one (M17). These mannitol derivatives might have similar interaction mode with HIV-1 protease. Then, 3D-QSAR was used to construct quantitative structure-activity models. The cross-validated q(2) values are found as 0.728 and 0.611 for CoMFA and CoMSIA, respectively. And the non-cross-validated r(2) values are 0.973 and 0.950. Nine test set compounds validate the model. The results show that this model possesses better prediction ability than the previous work. This model can be used to design new chemical entities and make quantitative prediction of the bioactivities for HIV-1 protease inhibitors before resorting to in vitro and in vivo experiment.

A comparison of different electrostatic potentials on prediction accuracy in CoMFA and CoMSIA studies

Computational chemistry is playing an increasingly important role in drug design and discovery, structural biology, and quantitative structure-activity relationship (QSAR) studies. For QSAR work, selecting an appropriate and accurate method to assign the electrostatic potentials of each atom in a molecule is a critical first step. So far several commonly used methods are available to assign charges. However, no systematic comparison of the effects of electrostatic potentials on QSAR quality has been made. In this study, twelve semi-empirical and empirical charge-assigning methods, AM1, AM1-BCC, CFF, Del-Re, Formal, Gasteiger, Gasteiger-Hückel, Hückel, MMFF, PRODRG, Pullman, and VC2003 charges, have been compared for their performances in CoMFA and CoMSIA modeling using several standard datasets. Some charge assignment models, such as Del-Re, PRODRG, and Pullman, are limited to specific atom and bond types, and, therefore, were excluded from this study. Among the remaining nine methods, the Gasteiger-Hückel charge, though commonly used, performed poorly in prediction accuracy. The AM1-BCC method was better than most charge-assigning methods based on prediction accuracy, though it was not successful in yielding overall higher cross-validation correlation coefficient (q(2)) values than others. The CFF charge model worked the best in prediction accuracy when q(2) was used as the evaluation criterion. The results presented should help the selection of electrostatic potential models in CoMFA and CoMSIA studies.

Comparison data sets for benchmarking QSAR methodologies in lead optimization

2D and 3D QSAR techniques are widely used in lead optimization-like processes. A compilation of 40 diverse data sets is described. It is proposed that these can be used as a common benchmark sample for comparisons of QSAR methodologies, primarily in terms of predictive ability. Use of this benchmark set will be useful for both assessment of new methods and for optimization of existing methods.

Partial charge calculation method affects CoMFA QSAR prediction accuracy

The 3D-QSAR method comparative molecular field analysis (CoMFA) involves the estimation of atomic partial charges as part of the process of calculating molecular electrostatic fields. Using 30 data sets from the literature the effect of using different common partial charge calculation methods on the predictivity (cross-validated R2) of CoMFA was studied. The partial charge methods ranged from the popular Gasteiger and the newer MMFF94 electronegativity equalization methods, to the more complex and computationally expensive semiempirical charges AM1, MNDO, and PM3. The MMFF94 and semiempirical MNDO, AM1, and PM3 methods for computing charges were found to result in statistically significantly more predictive CoMFA models than the Gasteiger charges. Although there was a trend toward the semiempirical charges performing better than the MMFF94 charges, the difference was not statistically significant. Thus, semiempirical partial charge calculation methods are suggested for the most predictive CoMFA models, but the MMFF94 charge calculation method is a very good alternative if semiempirical methods are not available or faster calculation speed is important.

Effect of steric molecular field settings on CoMFA predictivity

Steric molecular field can be represented in a number of ways in comparative molecular field analysis (CoMFA). This study aimed to investigate whether the choice of steric molecular field settings significantly influences the predictive performance of CoMFA and, if so, which is the best. The three-dimensional quantitative structure activity relationship (3D-QSAR) models based on Lennard-Jones, indicator, parabolic and Gaussian steric fields were compared using 28 datasets taken from the literature. The analysis of the predictive ability of these models (cross validated R(2)) indicates that steric fields in which the value drops off quickly with distance (i.e. Lennard-Jones and indicator fields) tend to perform better than the Gaussian version, which has a slower and smoother decrease. Furthermore, depending on the steric field type used, the field sampling density (i.e. grid spacing) has a variable influence on the predictive ability of the models generated.

Systematic Statistical Comparison of Comparative Molecular Similarity Indices Analysis Molecular Fields for Computer-Aided Lead Optimization

Comparative molecular similarity indices analysis (CoMSIA) is a 3D quantitative structure-activity relationship technique used to determine structural and electronic features influencing biological activity. This proves particularly useful for facilitating lead optimization projects. This study aimed to compare CoMSIA models produced using different subsets of the CoMSIA molecular fields (steric, electrostatic, hydrophobic, hydrogen-bond donor, and hydrogen-bond acceptor) in a systematic and statistically valid manner. A total of 23 data sets sourced from the literature were used to compare molecular field contribution and model predictivity using leave-one-out cross-validated R2 values. Predictive ability varied in a highly statistically significant manner depending on the set of CoMSIA molecular fields used. In general, the greater the number of CoMSIA molecular fields included in the analysis, the better the model predictivity was. There is great redundancy in the information contained in the different CoMSIA molecular fields. When all five CoMSIA molecular fields are included, the hydrophobic and electrostatic fields had the largest and the steric field the smallest contribution. Data sets were clustered into four groups on the basis of the utility of molecular field sets to generate predictive models.

Two- and three-dimensional quantitative structure–activity relationships for a series of purine nucleoside phosphorylase inhibitors

Comparative molecular field analysis (CoMFA), comparative molecular similarity indices analysis, and hologram quantitative structure-activity relationship (HQSAR) studies were conducted on a series of 52 training set inhibitors of calf spleen purine nucleoside phosphorylase (PNP). Significant cross-validated correlation coefficients (CoMFA, q(2)=0.68; CoMSIA, q(2)=0.66; and HQSAR, q(2)=0.70) were obtained, indicating the potential of the models for untested compounds. The models were then used to predict the inhibitory potency of 16 test set compounds that were not included in the training set, and the predicted values were in good agreement with the experimental results. The final QSAR models along with the information gathered from 3D contour and 2D contribution maps should be useful for the design of novel inhibitors of PNP having improved potency.

3D-QSAR Studies on Cannabinoid CB1 Receptor Agonists: G-Protein Activation as Biological Data

G-protein activation via the CB1 receptor was determined for a group of various CB1 ligands and utilized as biological activity data in subsequent CoMFA and CoMSIA studies. Both manual techniques and automated docking at CB1 receptor models were used to obtain a common alignment of endocannabinoid and classical cannabinoid derivatives. In the final alignment models, the endocannabinoid headgroup occupies a unique region distinct from the classical cannabinoid structures, supporting the hypothesis that these structurally diverse molecules overlap only partially within the receptor binding site. Both CoMFA and CoMSIA produce statistically significant models based on the manual alignment and a docking alignment at one receptor conformer. Leave-half-out cross-validation and progressive scrambling were successfully used in assessing the predictivity of the QSAR models.

BRUTUS: Optimization of a Grid-Based Similarity Function for Rigid-Body Molecular Superposition. 1. Alignment and Virtual Screening Applications

We have developed a fast grid-based algorithm, BRUTUS, for rigid-body molecular superposition and similarity searching. BRUTUS aligns molecules using field information derived from charge distributions and van der Waals shapes of the compounds. Molecules can have similar biological properties if their charge distributions and shapes are similar, even though they have different chemical structures; that is, BRUTUS can identify compounds possessing similar properties, regardless of their structures. In this paper, we present two applications of BRUTUS. First, BRUTUS was used to superimpose five sets of inhibitors. Second, two sets of known inhibitors were searched from a database, and the results were analyzed using self-organizing maps. We demonstrate that BRUTUS is successful in superimposing compounds using molecular fields and, importantly, is fast and accurate enough for virtual screening of chemical databases using a standard personal computer. This fast and efficient molecular-field-based algorithm is applicable for virtual screening of structurally diverse, active molecules.

Three-dimensional structure-activity relationships of nonsteroidal ligands in complex with androgen receptor ligand-binding domain

We studied the three-dimensional quantitative structure-activity relationships (3D QSAR) of 70 structurally and functionally diverse androgen receptor (AR) binding compounds using the comparative molecular similarity indices analysis (CoMSIA) method. The compound set contained 67 nonsteroidal analogues of flutamide, nilutamide, and bicalutamide whose binding mode to AR was unknown. Docking was used to identify the preferred binding modes for the nonsteroidal compounds within the AR ligand-binding pocket (LBP) and to generate the ligand alignment for the 3D QSAR analysis. The alignment produced a statistically significant and predictive model, validated by random group cross-validation and external test sets (q(2)(LOO) = 0.656, SDEP = 0.576, r(2) = 0.911, SEE = 0.293; q(2)(10) = 0.612, q(2)(5) = 0.571; pred-r(2) = 0.800). Additional model validation comes from the CoMSIA maps that were interpreted with respect to the LBP structure. The model takes into account and links the AR LBP structure, docked ligand structures, and the experimental binding activities. The results provide valuable information on intermolecular interactions between nonsteroidal ligands and the AR LBP.