External Predictive Ability Research Articles

Toward a unifying strategy for the structure-based prediction of toxicological endpoints.

Most computational methods used for the prediction of toxicity endpoints are based on the assumption that similar compounds have similar biological properties. This principle can be exploited using computational methods like read across or quantitative structure-activity relationships. However, there is no general agreement about which method is the most appropriate for quantifying compound similarity neither for exploiting the similarity principle in order to obtain reliable estimations of the compound properties. Moreover, optimal similarity metrics and modeling methods might depend on the characteristics of the endpoints and training series used in each case. This study describes a comparative analysis of the predictive performance of diverse similarity metrics and modeling methods in toxicological applications. A collection of two quantitative (n=660, n=1114) and three qualitative (n=447, n=905, n=1220) datasets representing very different endpoints of interest in drug safety evaluation and rigorous methods were used to estimate the external predictive ability in each case. The results confirm that no single approach produces the best results in all instances, and the best predictions were obtained using different tools in different situations. The trends observed in this study were exploited to propose a unifying strategy allowing the use of the most suitable method for every compound. A comparison of the quality of the predictions obtained by the unifying strategy with those obtained by standard prediction methods confirmed the usefulness of the proposed approach.

3D-QSAR analysis of MCD inhibitors by CoMFA and CoMSIA.

Three-dimensional quantitative structure-activity relationship was developed for the series of compounds as malonyl-CoA decarboxylase antagonists (MCD) using the CoMFA and CoMSIA methods. The statistical parameters for CoMFA (q(2)=0.558, r(2)=0.841) and CoMSIA (q(2)= 0.615, r(2) = 0.870) models were derived based on 38 compounds as training set in the basis of the selected alignment. The external predictive abilities of the built models were evaluated by using the test set of nine compounds. From obtained results, the CoMSIA method was found to have highly predictive capability in comparison with CoMFA method. Based on the given results by CoMSIA and CoMFA contour maps, some features that can enhance the activity of compounds as MCD antagonists were introduced and used to design new compounds with better inhibition activity.

Linear and non-linear modelling of the cytotoxicity of TiO2 and ZnO nanoparticles by empirical descriptors

Titanium oxide (TiO2) and zinc oxide (ZnO) nanoparticles are among the most widely used in different applications in daily life. In this study, local regression and classification models were developed for a set of ZnO and TiO2 nanoparticles tested at different concentrations for their ability to disrupt the lipid membrane in cells. Different regression techniques were applied and compared by checking the robustness of the models and their external predictive ability. Additionally, a simple classification model was developed, which predicts the potential for disruption of the studied nanoparticles with good accuracy (overall accuracy, specificity, and sensitivity >80%) on the basis of two empirical descriptors. The present study demonstrates that empirical descriptors, such as experimentally determined size and tested concentrations, are relevant to modelling the activity of nanoparticles. This information may be useful to screen the potential for harmful effect of nanoparticles in different experimental conditions and to optimize the design of toxicological tests. Results from the present study are useful to support and refine the future application of in silico tools to nanoparticles, for research and regulatory purposes.

Identification of 3-Nitro-2,4,6-trihydroxybenzamide Derivatives as Photosynthetic Electron Transport Inhibitors by QSAR and Pharmacophore Studies.

In the present investigation, quantitative structure-activity relationship (QSAR) analysis was performed on a data set consisting of structurally diverse compounds in order to investigate the role of their structural features on their photosynthetic electron transport Inhibitors. The best 2D-QSAR model was selected, having correlation coefficient r (2)=0.8544 and cross-validated squared correlation coefficient q (2)=0.7139 with external predictive ability of pred_r (2)=0.7753. The results obtained in this study indicate that the presence of hydroxy and nitro groups, expressed by the SsOHcount and SddsN (nitro) count, is the most relevant molecular property determining efficiency of photosynthetic inhibitory. Molecular field analysis was used to construct the best k-nearest neighbor (kNN-MFA)-based 3D-QSAR model using SA-PLS method, showing good correlative and predictive capabilities in terms of [Formula: see text] and [Formula: see text]. The pharmacophore model includes three features viz. hydrogen bond donor, hydrogen bond acceptor, and one aromatic feature. The developed model was found to be predictive and can be used to design potent photosynthetic electron transport activities prior to their synthesis for further lead modification.

Structural Requirements of Some 2-(1-Propylpiperidin-4-yl)-1H-benzimidazole-4-carboxamide Derivatives as Poly (ADP-Ribose) Polymerase (PARP) for the Treatment of Cancer: QSAR Approach.

The present study is aimed to elucidate the structural features of substituted 2-(1-propylpiperidin-4-yl)-1H-benzimidazole-4-carboxamide required for poly (ADP-ribose) polymerase inhibition and to obtain predictive 2D QSAR models to guide the rational synthesis of novel poly (ADP-ribose) polymerase inhibitors. The statistical analysis has shown that excellent results are obtained by using partial least regression based on simulated annealing method. The best model was selected based on the highest correlation coefficient r (2)=0.8590, and cross-validated squared correlation coefficient q (2)=0.7875 with external predictive ability of [Formula: see text] was developed by stepwise PLS method with the descriptors like T_N_F_1, SdsCHcount, and Rotatable Bond Count. The generated models provide insight into the influence of various interactive fields on the activity and, thus, can help in designing and forecasting the inhibition activity of novel (ADP-ribose) polymerase molecules.

A Structure–Activity Relationship Study of Naphthoquinone Derivatives as Antitubercular Agents Using Molecular Modeling Techniques

Tuberculosis (TB) is one of the major causes of death worldwide. Mycobacterium tuberculosis, the leading causative agent of TB, is responsible for the morbidity and mortality of a large population worldwide. In view of above and as a part of our effort to develop new and potent anti-TB agents, a series of substituted naphthoquinone derivatives were subjected to molecular modeling using various feature selection methods. The statistically significant best 2D-QSAR model having correlation coefficient [Formula: see text] and cross-validated squared correlation coefficient [Formula: see text] with external predictive ability of [Formula: see text] was developed by SA-PLS, and group-based QSAR model having [Formula: see text] and [Formula: see text] with [Formula: see text] was developed by SA-PLS. Further analysis using three-dimensional QSAR technique identifies a suitable model obtained by SA-partial least square method leading to antitubercular activity prediction. k-nearest neighbor molecular field analysis was used to construct the best 3D-QSAR model using SA-PLS method, showing good correlative and predictive capabilities in terms of [Formula: see text] and [Formula: see text]. The pharmacophore analysis results obtained from this study show that the distance between the aromatic/hydrophobic and the naphthoquinone moiety sites to the aliphatic and acceptor groups should be connected with almost the same distance for significant antitubercular activity. The information rendered by QSAR models may lead to a better understanding of structural requirements of antitubercular activity and also can help in the design of novel potent antitubercular activity.

QSAR studies on 3-(4-biphenylmethyl) 4, 5-dihydro-4-oxo-3H-imidazo [4, 5-c] Pyridine derivatives as angiotensin II (AT1) receptor antagonist

QSAR studies were performed for correlating the chemical composition of 3-(4-biphenylmethyl) 4, 5-dihydro-4-oxo-3H-imidazo [4, 5-c] pyridines bearing aryl acetic acid esters and acetamides as angiotensin II AT(1) receptor antagonist. Four different quantitative structure-property relationship (QSAR) methods namely two-dimensional (2D-QSAR), group-based QSAR, k-nearest neighbor and Pharmacophore Modeling were employed to obtain statistically significant models. The statistically significant best 2D-QSAR model having correlation coefficient r(2) = 0:8940 and cross-validated squared correlation coefficient q(2) = 0:7648 with external predictive ability of pred_r(2) = 0:8177,pred_r(2)se = 0.4119 and best group-based QSAR model having r(2) = 0:7392 and q(2) = 0:6710with pred_r(2) = 0:7503was developed by SA-principal component regression. The most predictive k-nearest neighbor model derived from the superposition of conformations has good cross-validated q(2) = 0:7637 and satisfied predictive ability r(2)_pred = 0.7143. Continuing with compounds of substituted 4, 5-dihydro-4-oxo-3H-imidazo [4, 5-c] pyridine derivatives chemical feature-based pharmacophore models with lowest RMSD value (0.3292 Å) consists of two Hac (Hydrogen bond acceptor), negative ionizable, and two AroC (Aromatic) features are important for the activity. The study suggested that substitution of group at R, R 1, R 2 and Ar, and position on 4, 5-dihydro-4-oxo-3H-imidazo [4, 5-c] pyridine ring with more electronegative nature and low bulkiness are favorable for the antihypertensive activity. These theoretical results may provide a useful reference for understanding the action mechanism and designing potential angiotensin II (AT1) receptor antagonist.

Nonlinear QSAR models with high-dimensional descriptor selection and SVR improve toxicity prediction and evaluation of phenols on Photobacterium phosphoreum

Assessment of the risk of chemicals is an important task in the environmental protection. In this paper, we developed quantitative structure–activity relationship (QSAR) methods to evaluate the toxicity of phenol to Photobacterium phosphoreum, which is an important indicator for water quality. We first built support vector regression (SVR) model using three descriptors, and the SVR model (t=2) had the highest external prediction ability (MSEext=0.068, Qext2=0.682), about 40% higher than literature model's. Second, to identify more effective descriptors, we applied in-house methods to select descriptors with clear meanings from 2835 descriptors calculated by the PCLIENT and used them to construct the SVR models. Our results showed that our twenty new QSAR models significantly increased the standard regression coefficient on test set (MSEext values ranged from 0.003 to 0.063 and Qext2 values ranged from 0.708 to 0.985). The Y random response permutation test and different splits of training/test datasets also supported the excellent predictive power of the best SVR model. We further evaluated the regression significance of our SVR model and the importance of each single descriptor of the model according to the interpretability analysis. Our work provided useful theoretical understanding of the toxicity of phenol analogues.

QSAR Studies of 3, 4-dihydropyrimidin-2(1H)-one Urea Derivatives as Antibacterial and Antifungal activity

The present study is aimed to elucidate the structural features of novel 3,4-dihydropyrimidin-2(1H)-one urea derivatives required for antibacterial and antifungal activity and to obtain predictive models to guide the rational synthesis of novel activity. The significant best 2D QSAR model (S.aureus) having correlation coefficient r2 = 0.7854 and crossvalidated squared correlation coefficient q2 = 0.7563 with external predictive ability of pred_r2 = 0.7629 was developed. The main aimed to obtain predictive models which may guide the synthesis of more active antimicrobial agents.

3D-QSAR, molecular docking and molecular dynamics studies of a series of RORγt inhibitors

The discovery of clinically relevant inhibitors of retinoic acid receptor-related orphan receptor-gamma-t (RORγt) for autoimmune diseases therapy has proven to be a challenging task. In the present work, to find out the structural features required for the inhibitory activity, we show for the first time a three-dimensional quantitative structure–activity relationship (3D-QSAR), molecular docking and molecular dynamics (MD) simulations for a series of novel thiazole/thiophene ketone amides with inhibitory activity at the RORγt receptor. The optimum CoMFA and CoMSIA models, derived from ligand-based superimposition I, exhibit leave-one-out cross-validated correlation coefficient (R2cv) of .859 and .805, respectively. Furthermore, the external predictive abilities of the models were evaluated by a test set, producing the predicted correlation coefficient (R2pred) of .7317 and .7097, respectively. In addition, molecular docking analysis was applied to explore the binding modes between the inhibitors and the receptor. MD simulation and MM/PBSA method were also employed to study the stability and rationality of the derived conformations, and the binding free energies in detail. The QSAR models and the results of molecular docking, MD simulation, binding free energies corroborate well with each other and further provide insights regarding the development of novel RORγt inhibitors with better activity.

Structural requirements of some 2-(1-propylpiperidin-4-yl)-1H-benzimidazole-4-carboxamide derivatives as poly (ADP-ribose) polymerase (PARP) for the treatment of cancer: QSAR approach.

The present study is aimed to elucidate the structural features of substituted 2-(1-propylpiperidin-4-yl)-1H-benzimidazole-4-carboxamide required for poly (ADP-ribose) polymerase inhibition and to obtain predictive 2D QSAR models to guide the rational synthesis of novel poly (ADP-ribose) polymerase inhibitors. The statistical analysis has shown that excellent results are obtained by using partial least regression based on simulated annealing method. The best model was selected based on the highest correlation coefficient r2 = 0.8590 and cross validated squared correlation coefficient q2 = 0.7875 with external predictive ability of pred_r2 = 0.7407 was developed by stepwise PLS method with the descriptors like T_N_F_1, SdsCHcount, and Rotatable Bond Count. The generated models provide insight into the influence of various interactive fields on the activity and, thus, can help in designing and forecasting the inhibition activity of novel (ADP-ribose) polymerase molecules.

Identification of 3-nitro-2, 4, 6-trihydroxybenzamide derivatives as photosynthetic electron transport inhibitors by QSAR and pharmacophore studies.

In the present investigation, QSAR analysis was performed on a data set consist of structurally diverse compounds in order to investigate the role of its structural features on their Photosynthetic Electron Transport Inhibitors. The herbicidal activity co-related with certain topological and hydrophobicity based descriptors, 3D descriptors dependent steric, electrostatic and hydrophobic. The best 2D QSAR model was selected, having correlation coefficient r2 = 0.8544 and cross validated squared correlation coefficient q2 = 0.7139 with external predictive ability of pred_r2 = 0.7753 was developed. The results obtained in this study indicate that hydroxy and nitro groups, as expressed by the SsOHcount, SddsN (nitro) count, is the most relevant molecular property determining efficiency of photosynthetic inhibitory. Molecular field analysis was used to construct the best k-nearest neighbor (kNN-MFA)-based 3DQSAR model using SA-PLS method, showing good correlative and predictive capabilities in terms of q2 = 0.7694 and pred_r2 = 0.7381. The influences of steric, electrostatic and hydrophobic field effects generated by the contribution plots are discussed. The pharmacophore model includes three features viz. hydrogen bond donor, hydrogen bond acceptor, and one aromatic feature was developed. The developed model was found to be predictive and can be used to design potent Photosynthetic Electron Transport activities prior to their synthesis for further lead modification. The results obtained suggest that the 3-nitro-2, 4, 6-trihydroxybenzamide analogues represent promising candidates for the development of new active principles targeting photosynthesis to be used as herbicides.

In silico study combining docking and QSAR methods on a series of matrix metalloproteinase 13 inhibitors.

Matrix metalloproteinase 13 (MMP-13) plays an important role in the degradation of articular cartilage and has been considered as an attractive target for the treatment of osteoarthritis; hence, the development of efficient inhibitors of MMP-13 has become a hot study field. Taking a series of carboxylic acid-based MMP-13 inhibitors as research object, this work utilized an extended QSAR method to analyze the structure-activity relationships. We focused on two important topics in QSAR: bioactive conformation and descriptors. Firstly, molecular docking was carried out to dock all molecules into the MMP-13 active site in order to obtain the bioactive conformation. Secondly, based on the docked complex, descriptors characterizing receptor-ligand interactions and the ligand structure were calculated. Thirdly, a genetic algorithm (GA) and multiple linear regression (MLR) were employed to select important descriptors related to inhibitory activities, simultaneously, to build the predictive model. The built model gave satisfactory results with highly accurate fitting and strong external predictive abilities for chemicals not used in model development. Furthermore, the selected descriptors were explored to elucidate important factors influencing the inhibition activities. This study demonstrates that the selection strategy of the docking-guided bioactive conformation is rational and useful in predicting MMP-13 inhibitor activities, and receptor-ligand complex descriptors have an advantage over directly reflecting receptor-ligand interactions.

Molecular modeling studies on substituted aminopyrimidines derivatives as potential antimalarial compounds

We report combined study of k-nearest neighbor, pharmacophore and 2D QSAR was performed on a series of 2,4-diaminopyrimidines dihydrofolate reductase from pyrimethamine-resistant Plasmodium falciparum as antimalarial agents to gain insights into the structural determinants and their structure–activity relationship. The QSAR models for the prediction of activity of antiplasmodial activities against P. falciparum clones with wild-type TM4/8.2 and K1CB1 strains have been developed by the SA-PLS and SW-PLS methods, and the proposed models gain satisfactory results. The statistically significant best 2D QSAR model having correlation coefficient r2 = 0.8569, 0.7853 and cross-validated squared correlation coefficient q2 = 0.7104, 0.7216 with external predictive ability of pred_r2 = 0.7995, 0.7064 was developed by wild-type TM4/8.2 and K1CB1 strains with SA-PLS. 3D QSAR studies using k-nearest neighbor molecular field analysis (kNN-MFA) method, identifies two models obtained by SA-PLS and SW-PLS methods leading to antimalarial activity prediction. The obtained pharmacophore model with lowest RMSD value (0.1548 A), consisting of one hydrogen donor, two hydrogen acceptors and one aromatic region was developed. These models were found to yield reliable clues for further optimization of 2,4-diaminopyrimidines derivatives in the data set. We hope that these results will give new insights into chemical modifications that can be realized with the aim of designing new inhibitors with improved pharmacological properties.

Effect of information leakage and method of splitting (rational and random) on external predictive ability and behavior of different statistical parameters of QSAR model

Quantitative Structure–Activity Relationship not only provides guidelines regarding structural features responsible for biological activity but it can be used also for prediction of desired activity prior to synthesis of untested chemicals. Therefore, an appropriate validation of any QSAR is of utmost importance to judge its external predictive ability. Generally, internal and external validations (preferred by many) are used in the absence of a true external dataset. The model developed using external method may not be reliable as it may not capture all essential features required for the particular SAR due to omission of some compounds, especially for small datasets. In external validation, the splitting is done either rationally or in random manner before descriptor selection. In the present study, rational splitting of dataset was performed using a novel method and its effect on statistical parameters was analyzed. The analysis reveals that the predictive ability of a QSAR model is sensitive toward (1) the method of splitting and (2) distribution of the training and the prediction sets. In addition, purposeful selection can be used to influence the statistical parameters; therefore, external validation based on single split is insufficient to guarantee the true predictive ability of a QSAR model. Besides, it appears that the selection of descriptors prior to splitting (information leakage) has little role to play in deciding external predictivity of the model. The present study reveals that as many as possible statistical parameters should be examined along with boot-strapping instead of single external validation.

QSAR of antitrypanosomal activities of polyphenols and their analogues using multiple linear regression and artificial neural networks.

In order to find a thriving quantitative structure-activity relationship for antitrypanosomal activities (against Trypanosma brucei rhodesiense) of polyphenols that belong to different structural groups, multiple linear regression (MLR) and artificial neural networks (ANN) were employed. The analysis was performed on two different-sized training sets (59% and 78% molecules in the training set), resulting in relatively successful MLR and ANN models for the data set containing the smaller training set. The best MLR model obtained using the five descriptors (R3m(+), GAP, DISPv, HATS2m, JGI2) was able to account only for 74% of the variance of antitrypanosomal activities of the training set and achieved a high internal, but low external prediction. Nonlinearities of the best ANN model compared with the linear model improved the coefficient of determination to 98.6%, and showed a better external predictive ability. The obtained models displayed relevance of the distance between oxygen atoms in molecules of polyphenols, as well as stability of molecules, measured by the difference between the energy of the highest occupied molecular orbital and the energy of the lowest unoccupied molecular orbital (GAP) for their activity.

Pharmacophore and docking-based virtual screening approach for the design of new dual inhibitors of Janus kinase 1 and Janus kinase 2

Janus kinase 1 and 2, non-receptor protein tyrosine kinases, are implicated in various cancerous diseases. Involvement of these two enzymes in the pathways that stimulate cell proliferation in cancerous conditions makes them potential therapeutic targets for designing new dual-targeted agents for the treatment of cancer. In the present study, two separate pharmacophore models were developed and the best models for JAK1 (AAADH.25) and JAK2 (ADRR.92) were selected on the basis of their external predictive ability. Both models were subjected to a systematic virtual screening (VS) protocol using a PHASE database of 1.5 million molecules. The hits retrieved in VS were investigated for ADME properties to avoid selection of molecules with a poor pharmacokinetic profile. The molecules considered to be within the range of acceptable limits of ADME properties were further employed for docking simulations with JAK1 and JAK2 proteins to explore the final hits that possess structural features of both pharmacophore models as well as display essential interactions with both of them. Thus, the new molecules obtained in this way should show inhibitory activity against JAK1 and JAK2 and may serve as novel therapeutic agents for the treatment of cancerous disease conditions.

Are bigger data sets better for machine learning? Fusing single-point and dual-event dose response data for Mycobacterium tuberculosis.

Tuberculosis is a major, neglected disease for which the quest to find new treatments continues. There is an abundance of data from large phenotypic screens in the public domain against Mycobacterium tuberculosis (Mtb). Since machine learning methods can learn from past data, we were interested in addressing whether more data builds better models. We now describe using Bayesian machine learning to assess whether we can improve our models by combining the large quantities of single-point data with the much smaller (higher quality) dual-event data sets, which use both dose-response data for both whole-cell antitubercular activity and Vero cell cytotoxicity. We have evaluated 12 models ranging from different single-point, dual-event dose-response, single-point and dual-event dose-response as well as combined data sets for three distinct data sets from the same laboratory. We used a fourth data set of active and inactive compounds from the same group as well as a smaller set of 177 active compounds from GlaxoSmithKline as test sets. Our data suggest combining single-point with dual-event dose-response data does not diminish the internal or external predictive ability of the models based on the receiver operator curve (ROC) for these models (internal ROC range 0.83-0.91, external ROC range 0.62-0.83) compared to the orders of magnitude smaller dual-event models (internal ROC range 0.6-0.83 and external ROC 0.54-0.83). In conclusion, models developed with 1200-5000 compounds appear to be as predictive as those generated with 25 000-350 000 molecules. Our results have implications for justifying further high-throughput screening versus focused testing based on model predictions.

Induced fit docking, pharmacophore modeling, and molecular dynamic simulations on thiazolidinedione derivatives to explore key interactions with Tyr48 in polyol pathway.

To obtain a scientific thought and expedition to explore key interactions with Tyr48 in aldose reductase (ALR), combined study of pharmacophore modeling, induced fit docking, and dynamics studies were performed on ALR. A stereo chemically and energetically valid model of ALR-NADP+ complex was developed using homology modeling technique. Statistically a significant five point pharmacophore model was designed on a set of 54 thiazolidinedione derivatives with good external and internal predictive ability. Rigid and induced fit docking protocols were applied on ALR protein for both with and without NADP+ cofactor to identify a suitable binding mode that facilitates the key hydrogen bond interactions with Tyr48. Docking of thiazolidinedione derivatives into ALR-NADP+ complex gave more promising results by reducing false positive binding of inhibitors into the co-factor binding site. Structural changes within Try48 and Asp43 during the binding process in enzyme inhibitor complex were analyzed using molecular dynamics (MD) simulations. The results obtained from dynamic simulations emphasized the role of Tyr48 in maintaining inter or intra molecular hydrogen bond interaction with the protein or inhibitor respectively. New molecules were designed and checked for their binding interactions and showed improved results compared to existing thiazolidinediones derivatives. Hence, these combined protocols will be helpful and cooperative to design and optimize molecules with better inhibitory activity against the biologically active target.

Topomer CoMFA and Virtual Screening Studies of Azaindole Class Renin Inhibitors

Direct renin inhibitors (DRIs) have increasingly shown a significant advantage in the treatment of hypertension and protection of target organs. In this paper, a series of azaindole class renin inhibitors were subjected to 3D-QSAR study using Topomer CoMFA. Five kinds of splitting mode for different fragment cutting and 6 different training and test sets grouping were attempted to build consensus models. The results indicated that 6 consensus models had similar predictability (q(2) and r(2) pred) and stability (r(2)). The best model showed good stability and predictability (q(2) of 0.616, r(2) of 0.908). The r(2) pred value of the external test set was 0.70, which means that the model had an excellent external predictive ability, and the robustness of the developed model was assessed by the Y-randomization test. This study also adopted the methodology of fragment-based drug design (FBDD) to virtual screen new renin inhibitors by using Topomer Search technology. The R1-group of the compound No. 13 with the highest activity was chosen as the basic scaffold, and its remaining R2-group acted as a query to screen 142,025 molecules of ZINC database for similar fragments. The obtained 30 fragments with the highest R2-group contribution values were added to the basic scaffold respectively. Finally 30 new azaindole compounds with potent high activities were obtained. Further the binding modes were studied by using Surflex- Dock. The docking results showed good binding interactions of the designed compounds with the renin protein, thus the rationality of this design was further verified from the perspective of the renin receptor.