Vasyl Kovalishyn

Neural Network Studies. 3. Variable Selection in the Cascade-Correlation Learning Architecture

Pruning methods for feed-forward artificial neural networks trained by the cascade-correlation learning algorithm are proposed. The cascade-correlation algorithm starts with a small network and dynamically adds new nodes until the analyzed problem has been solved. This feature of the algorithm removes the requirement to predefine the architecture of the neural network prior to network training. The developed pruning methods are used to estimate the importance of large sets of initial variables for quantitative structure−activity relationship studies and simulated data sets. The calculated results are compared with the performance of fixed-size back-propagation neural networks and multiple regression analysis and are carefully validated using different training/test set protocols, such as leave-one-out and full cross-validation procedures. The results suggest that the pruning methods can be successfully used to optimize the set of variables for the cascade-correlation learning algorithm neural networks. Th...

Antibacterial Activity of Imidazolium-Based Ionic Liquids Investigated by QSAR Modeling and Experimental Studies

Predictive QSAR models for the inhibitors of B. subtilis and Ps. aeruginosa among imidazolium‐based ionic liquids were developed using literary data. The regression QSAR models were created through Artificial Neural Network and k‐nearest neighbor procedures. The classification QSAR models were constructed using WEKA‐RF (random forest) method. The predictive ability of the models was tested by fivefold cross‐validation; giving q2 = 0.77–0.92 for regression models and accuracy 83–88% for classification models. Twenty synthesized samples of 1,3‐dialkylimidazolium ionic liquids with predictive value of activity level of antimicrobial potential were evaluated. For all asymmetric 1,3‐dialkylimidazolium ionic liquids, only compounds containing at least one radical with alkyl chain length of 12 carbon atoms showed high antibacterial activity. However, the activity of symmetric 1,3‐dialkylimidazolium salts was found to have opposite relationship with the length of aliphatic radical being maximum for compounds based on 1,3‐dioctylimidazolium cation. The obtained experimental results suggested that the application of classification QSAR models is more accurate for the prediction of activity of new imidazolium‐based ILs as potential antibacterials.

The structure--antituberculosis activity relationships study in a series of 5-(4-aminophenyl)-4-substituted-2,4-dihydro-3h-1,2,4-triazole-3-thione derivatives. A combined electronic-topological and neural networks approach.

Antituberculosis activity of several 5-(4-aminophenyl)-4-alkyl/aryl-2,4-dihydro-3H-1,2,4-triazole-3-thiones (1-9) and their thiourea derivatives (10-31) were screened for their antimycobacterial activities against Mycobacterium tuberculosis H37Rv using the BACTEC 460 radiometric system. Of the synthesized compounds, 10-12, 30 were the most active derivatives exhibiting more than 90 % inhibition of mycobacterial growth at 12.5 microg/mL. Structure-activity relationships study was performed for the given series by using the Electronic-Topological Method combined with Neural Networks (ETM-NN). A system of prognosis was developed as the result of training associative neural network (ASNN) using weights of pharmacophoric fragments as descriptors. Descriptors were calculated by the projection of ETM compound and pharmacophoric fragments on the elements of Kohonens self-organizing maps (SOM). From the detailed analysis of all compounds under study, the necessary requirements for a compound to possess antituberculosis activity were formulated. The analysis have shown that any requirements violation for a molecule implies a considerable decrease or even complete loss of its activity.

Variable Selection in the Cascade-Correlation Learning Architecture

Recently there has been a growing interest in the application of neural networks in the field of QSAR. It was demonstrated that this method is often superior to the traditional approaches.1 Other studies have shown that prediction ability of such methods can be substantially improved if the number of input variables for neural networks is optimized.2,3

Design and Synthesis of New Potent Inhibitors of Farnesyl Pyrophosphate Synthase

Predictive QSAR models for the inhibition activities of nitrogen-containing bisphosphonates (N-BPs) against farnesyl pyrophosphate synthase (FPPS) from Leishmania major (LeFPPS) were developed using a data set of 97 compounds. The QSAR models were developed through the use of Artificial Neural Networks and Random Forest learning procedures. The predictive ability of the models was tested by means of leave-one-out cross-validation; Q(2)values ranging from 0.45-0.79 were obtained for the regression models. The consensus prediction for the external evaluation set afforded high predictive power (Q(2)=0.76 for 35 compounds). The robustness of the QSAR models was also evaluated using a Y-randomization procedure. A small set of 6 new N-BPs were designed and synthesized applying the Michael reaction of tetrakis (trimethylsilyl) ethenylidene bisphosphonate with amines. The inhibition activities of these compounds against LeFPPS were predicted by the developed QSAR models and were found to correlate with their fungistatic activities against Candida albicans. The antifungal activities of N-BPs bearing n-butyl and cyclopropyl side chains exceeded the activities of Fluconazole, a triazole-containing antifungal drug. In conclusion, the N-BPs developed here present promising candidate drugs for the treatment of fungal diseases.

Imidazolium Ionic Liquids as Potential Anti-Candida Inhibitors: QSAR Modeling and Experimental Studies

Quantitative structure-activity relationships (QSAR) of imidazolium ionic liquids (ILs) as inhibitors of C. albicans collection strains (IOA-109, KCTC 1940, ATCC 10231) have been studied. Predictive QSAR models were built using different descriptor sets for a set of 88 ionic liquids with known minimum inhibitory concentrations (MIC) against C. albicans. We applied the state-of-the-art QSAR methodologies such as WEKA Random Forest (RF) as a binary classifier, Associative Neural Networks (ASNN) and k-Nearest Neighbors (k-NN) to build continuum non-linear regression models. The obtained models were validated using a 5-fold cross-validation approach and resulted in the prediction accuracies of 80% ± 5.0 for the classification models and q2 = 0.73-0.87 for the non-linear regression models. Biological testing of newly synthesized 1,3-dialkylimidazolium ionic liquids with predicted activity was performed by disco-diffusion method against C. albicans ATCC 10231 M885 strain and clinical isolates C. albicans, C. krusei and C. glabrata strains. The high percentage of coincidence between the QSAR predictions and the experimental results confirmed the high predictive power of the developed QSAR models within the applicability domain of new imidazolium ionic liquids.

QSAR studies, synthesis and antibacterial assessment of new inhibitors against multidrug-resistant Mycobacterium tuberculosis.

BACKGROUND The increasing rate of appearance of multidrug-resistant strains of Mycobacterium tuberculosis (MDR-TB) is a serious problem at the present time. MDR-TB forms do not respond to the standard treatment with the commonly used drugs and can take some years or more to treat with drugs that are less potent, more toxic and much more expensive. OBJECTIVE The goal of this work is to identify the novel effective drug candidates active against MDR-TB strains through the use of methods of cheminformatics and computeraided drug design. METHODS This paper describes Quantitative Structure-Activity Relationships (QSAR) studies using Artificial Neural Networks, synthesis and in vitro antitubercular activity of several potent compounds against H37Rv and resistant Mycobacterium tuberculosis (Mtb) strains. RESULTS Eight QSAR models were built using various types of descriptors with four publicly available structurally diverse datasets, including recent data from PubChem and ChEMBL. The predictive power of the obtained QSAR models was evaluated with a cross-validation procedure, giving a q2=0.74-0.78 for regression models and overall accuracy 78.9-94.4% for classification models. The external test sets were predicted with accuracies in the range of 84.1-95.0% (for the active/inactive classifications) and q2=0.80- 0.83 for regressions. The 15 synthesized compounds showed inhibitory activity against H37Rv strain whereas the compounds 1-7 were also active against resistant Mtb strain (resistant to isoniazid and rifampicin). CONCLUSION The results indicated that compounds 1-7 could serve as promising leads for further optimization as novel antibacterial inhibitors, in particular, for the treatment of drug resistance of Mtb forms.

The Quantum Chemical and QSAR Studies on Acinetobacter Baumannii Oxphos Inhibitors

BACKGROUND Acinetobacter is a Gram-negative, catalase-positive, oxidase-negative, non-motile, and no fermenting bacteria. OBJECTIVE In this study, some of the electronic and molecular properties, such as the highest occupied molecular orbital energy (EHOMO), lowest unoccupied molecular orbital energy (ELUMO), the energy gap between EHOMO and ELUMO, Mulliken atomic charges, bond lengths, of molecules having impact on antibacterial activity against A. baumannii were studied. In addition, calculations of some QSAR descriptors such as global hardness, softness, electronegativity, chemical potential, global electrophilicity, nucleofugality, electrofugality were performed. METHOD The descriptors having impact on antibacterial activity against A. baumannii have been investigated based on the usage of 29 compounds employing two statistical methods called Linear Regression and Artificial Neural Networks. RESULTS Artificial Neural Networks obtained accuracies in the range of 83-100% (for active/inactive classifications) and q2=0.63 for regression. CONCLUSION Three ANN models were built using various types of descriptors with publicly available structurally diverse data set. QSAR methodologies used Artificial Neural Networks. The predictive ability of the models was tested with cross-validation procedure, giving a q2=0.62 for regression model and overall accuracy 70-95 % for classification models.

Design, synthesis and evaluation of novel sulfonamides as potential anticancer agents

Based on modern literature data about biological activity of E7010 derivatives, a series of new sulfonamides as potential anticancer drugs were rationally designed by QSAR modeling methods Сlassification learning QSAR models to predict the tubulin polymerization inhibition activity of novel sulfonamides as potential anticancer agents were created using the Online Chemical Modeling Environment (OCHEM) and are freely available online on OCHEM server at https://ochem.eu/article/107790. A series of sulfonamides with predicted activity were synthesized and tested against 60 human cancer cell lines with growth inhibition percent values. The highest antiproliferative activity against leukemia (cell lines K-562 and MOLT-4), non-small cell lung cancer (cell line NCI-H522), colon cancer (cell lines NT29 and SW-620), melanoma (cell lines MALME-3M and UACC-257), ovarian cancer (cell lines IGROV1 and OVCAR-3), renal cancer (cell lines ACHN and UO-31), breast cancer (cell line T-47D) was found for compounds 4-9. According to the docking results the compounds 4-9 induce cytotoxicity by the disruption of the microtubule dynamics by inhibiting tubulin polymerization via effective binding into colchicine domain, similar the E7010.

Rational design of isonicotinic acid hydrazide derivatives with antitubercular activity: Machine learning, molecular docking, synthesis and biological testing

The problem of designing new antitubercular drugs against multiple drug‐resistant tuberculosis (MDR‐TB) was addressed using advanced machine learning methods. As there are only few published measurements against MDR‐TB, we collected a large literature data set and developed models against the non‐resistant H37Rv strain. The predictive accuracy of these models had a coefficient of determination q2 = .7–.8 (regression models) and balanced accuracies of about 80% (classification models) with cross‐validation and independent test sets. The models were applied to screen a virtual chemical library, which was designed to have MDR‐TB activity. The seven most promising compounds were identified, synthesized and tested. All of them showed activity against the H37Rv strain, and three molecules demonstrated activity against the MDR‐TB strain. The docking analysis indicated that the discovered molecules could bind enoyl reductase, InhA, which is required in mycobacterial cell wall development. The models are freely available online (http://ochem.eu/article/103868) and can be used to predict potential anti‐TB activity of new chemicals.