Prediction of biological activity of Aurora-A kinase inhibitors by multilinear regression analysis and support vector machine

Several QSAR (quantitative structure-activity relationships) models for predicting the inhibitory activity of 117 Aurora-A kinase inhibitors were developed. The whole dataset was split into a training set and a test set based on two different methods, (1) by a random selection; and (2) on the basis of a Kohonen’s self-organizing map (SOM). Then the inhibitory activity of 117 Aurora-A kinase inhibitors was predicted using multilinear regression (MLR) analysis and support vector machine (SVM) methods, respectively. For the two MLR models and the two SVM models, for the test sets, the correlation coefficients of over 0.92 were achieved.     

Pharmacophore mapping of diverse classes of farnesyltransferase inhibitors

Protein farnesyltransferase (FTase) is a zinc-dependent enzyme that catalyzes the attachment of a farnesyl lipid group to the sulfur atom of a cysteine residue of numerous proteins involved in cell signaling including the oncogenic H-Ras protein. Pharmacophore models were developed by using Catalyst HypoGen program with a training set of 22 farnesyltransferase inhibitors (FTIs), which were carefully selected with great diversity in both molecular structure and bioactivity for discovering new potent FTIs. The best pharmacophore hypothesis (Hypo 1), consisting of four features, namely, one hydrogen-bond acceptor (HBA), one hydrophobic point (HY), and two ring aromatics (RA), has a correlation coefficient of 0.961, a root mean square deviation (RMSD) of 0.885, and a cost difference of 62.436, suggesting that a highly predictive pharmacophore model was successfully obtained. For the test series, a classification scheme was used to distinguish highly active from moderately active and inactive compounds on the basis of activity ranges. Hypo 1 was validated with 181 test set compounds, which has a correlation coefficient of 0.713 between estimated activity and experimentally measured activity. The model was further validated by screening a database spiked with 25 known inhibitors. The model picked up all 25 known inhibitors giving an enrichment factor of 10.892. The results demonstrate that the hypothesis derived in this study can be considered to be a useful and reliable tool in identifying structurally diverse compounds with desired biological activity.     

Molecular docking and 3D-QSAR studies on inhibitors of DNA damage signaling enzyme human PARP-1

Poly (ADP-ribose) polymerase-1 (PARP-1) operates in a DNA damage signaling network. Molecular docking and three dimensional-quantitative structure activity relationship (3D-QSAR) studies were performed on human PARP-1 inhibitors. Docked conformation obtained for each molecule was used as such for 3D-QSAR analysis. Molecules were divided into a training set and a test set randomly in four different ways, partial least square analysis was performed to obtain QSAR models using the comparative molecular field analysis (CoMFA) and comparative molecular similarity indices analysis (CoMSIA). Derived models showed good statistical reliability that is evident from their r2, q2(loo) and r2(pred) values. To obtain a consensus for predictive ability from all the models, average regression coefficient r2(avg) was calculated. CoMFA and CoMSIA models showed a value of 0.930 and 0.936, respectively. Information obtained from the best 3D-QSAR model was applied for optimization of lead molecule and design of novel potential inhibitors.     

3D-QSAR and molecular docking studies of 2-pyrimidinecarbonitrile derivatives as inhibitors against falcipain-3

The three dimensional-quantitative structure activity relationship (3D-QSAR) studies were performed on a series of falcipain-3 inhibitors using comparative molecular field analysis (CoMFA) and comparative molecular similarity indices analysis (CoMSIA) techniques. A training set containing 42 molecules served to establish the QSAR models. The optimum CoMFA and CoMSIA models obtained for the training set were statistically significant with cross-validated correlation coefficients r(cv)(2) (q(2)) of 0.549 and 0.608, and conventional correlation coefficients (r(2)) of 0.976 and 0.932, respectively. An independent test set of 12 molecules validated the external predictive power of both models with predicted correlation coefficients (r(pred)(2)) for CoMFA and CoMSIA as 0.697 and 0.509, respectively. The docking of inhibitors into falcipain-3 active site using GOLD software revealed the vital interactions and binding conformation of the inhibitors. The CoMFA and CoMSIA field contour maps agree well with the structural characteristics of the binding pocket of falcipain-3 active site, which suggests that the information rendered by 3D-QSAR models and the docking interactions can provide guidelines for the development of improved falcipain-3 inhibitors.     

Topological models for prediction of anti-inflammatory activity of N-arylanthranilic acids

Relationship of anti-inflammatory activity of N-arylanthranilic acids with distance based Wiener's index, adjacency based Zagreb indices M1 and M2, and distance-cum-adjacency based eccentric connectivity index (ECI) was investigated. A dataset comprising of 112 N-arylanthranilic acids was selected. The values of all the four indices for each of the 112 compounds were calculated using an in-house computer program. The dataset was divided randomly into training and test sets. The data was analyzed and suitable models were developed after identification the active ranges in the training set. Subsequently, a biological activity was assigned to each of the compound involved in the test set using these models, which was then compared with the reported anti-inflammatory activity. High accuracy of prediction ranging from 83% to 90% was observed using models based upon topological indices.     

Structure-based approach to pharmacophore identification, in silico screening, and three-dimensional quantitative structure-activity relationship studies for inhibitors of Trypanosoma cruzi dihydrofolate reductase function

We have employed a structure-based three-dimensional quantitative structure-activity relationship (3D-QSAR) approach to predict the biochemical activity for inhibitors of T. cruzi dihydrofolate reductase-thymidylate synthase (DHFR-TS). Crystal structures of complexes of the enzyme with eight different inhibitors of the DHFR activity together with the structure in the substrate-free state (DHFR domain) were used to validate and refine docking poses of ligands that constitute likely active conformations. Structural information from these complexes formed the basis for the structure-based alignment used as input for the QSAR study. Contrary to indirect ligand-based approaches the strategy described here employs a direct receptor-based approach. The goal is to generate a library of selective lead inhibitors for further development as antiparasitic agents. 3D-QSAR models were obtained for T. cruzi DHFR-TS (30 inhibitors in learning set) and human DHFR (36 inhibitors in learning set) that show a very good agreement between experimental and predicted enzyme inhibition data. For crossvalidation of the QSAR model(s), we have used the 10% leave-one-out method. The derived 3D-QSAR models were tested against a few selected compounds (a small test set of six inhibitors for each enzyme) with known activity, which were not part of the learning set, and the quality of prediction of the initial 3D-QSAR models demonstrated that such studies are feasible. Further refinement of the models through integration of additional activity data and optimization of reliable docking poses is expected to lead to an improved predictive ability.     

In silico approaches to identify novel myeloid cell leukemia-1 (Mcl-1) inhibitors for treatment of cancer

Myeloid cell leukemia-1 (Mcl-1) has been a validated and attractive target for cancer therapy. Over-expression of Mcl-1 in many cancers allows cancer cells to evade apoptosis and contributes to the resistance to current chemotherapeutics. Here, we identified new Mcl-1 inhibitors using a multi-step virtual screening approach. First, based on two different ligand-receptor complexes, 20 pharmacophore models were established by simultaneously using ‘Receptor-Ligand Pharmacophore Generation’ method and manual build feature method, and then carefully validated by a test database. Then, pharmacophore-based virtual screening (PB-VS) could be performed by using the 20 pharmacophore models. In addition, docking study was used to predict the possible binding poses of compounds, and the docking parameters were optimized before performing docking-based virtual screening (DB-VS). Moreover, a 3D QSAR model was established by applying the 55 aligned Mcl-1 inhibitors. The 55 inhibitors sharing the same scaffold were docked into the Mcl-1 active site before alignment, then the inhibitors with possible binding conformations were aligned. For the training set, the 3D QSAR model gave a correlation coefficient r² of 0.996; for the test set, the correlation coefficient r² was 0.812. Therefore, the developed 3D QSAR model was a good model, which could be applied for carrying out 3D QSAR-based virtual screening (QSARD-VS). After the above three virtual screening methods orderly filtering, 23 potential inhibitors with novel scaffolds were identified. Furthermore, we have discussed in detail the mapping results of two potent compounds onto pharmacophore models, 3D QSAR model, and the interactions between the compounds and active site residues.     

Support vector machines in HTS data mining: Type I MetAPs inhibition study

This article reports a successful application of support vector machines (SVMs) in mining high-throughput screening (HTS) data of a type I methionine aminopeptidases (MetAPs) inhibition study. A library with 43,736 small organic molecules was used in the study, and 1355 compounds in the library with 40% or higher inhibition activity were considered as active. The data set was randomly split into a training set and a test set (3:1 ratio). The authors were able to rank compounds in the test set using their decision values predicted by SVM models that were built on the training set. They defined a novel score PT50, the percentage of the test set needed to be screened to recover 50% of the actives, to measure the performance of the models. With carefully selected parameters, SVM models increased the hit rates significantly, and 50% of the active compounds could be recovered by screening just 7% of the test set. The authors found that the size of the training set played a significant role in the performance of the models. A training set with 10,000 member compounds is likely the minimum size required to build a model with reasonable predictive power.     

3D QSAR studies,pharmacophore modeling,and virtual screening of diarylpyrazole–benzenesulfonamide derivatives as a template to obtain new inhibitors,using human carbonic anhydrase II as a model protein

A 3D-QSAR modeling was performed on a series of diarylpyrazole-benzenesulfonamide derivatives acting as inhibitors of the metalloenzyme carbonic anhydrase (CA, EC 4.2.1.1). The compounds were collected from two datasets with the same scaffold, and utilized as a template for a new pharmacophore model to screen the ZINC database of commercially available derivatives. The datasets were divided into training, test, and validation sets. As the first step, comparative molecular field analysis (CoMFA), CoMFA region focusing and comparative molecular similarity indices analysis (CoMSIA) in parallel with docking studies were applied to a set of 41 human (h) CA II inhibitors. The validity and the prediction capacity of the resulting models were evaluated by leave-one-out (LOO) cross-validation approach. The reliability of the model for the prediction of possibly new CA inhibitors was also tested.     

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.