Use of Artificial Neural Network for a QSAR Study on Neurotrophic Activities of N-p-Tolyl/phenylsulfonyl L-Amino Acid Thiolester Derivatives

Diptool-A Novel Numerical Tool for Membrane Interactions Analysis, Applying to Antimicrobial Detergents and Drug Delivery Aids

The widespread problem of resistance development in bacteria has become a critical issue for modern medicine. To limit that phenomenon, many compounds have been extensively studied. Among them were derivatives of available drugs, but also alternative novel detergents such as Gemini surfactants. Over the last decade, they have been massively synthesized and studied to obtain the most effective antimicrobial agents, as well as the most selective aids for nanoparticles drug delivery. Various protocols and distinct bacterial strains used in Minimal Inhibitory Concentration experimental studies prevented performance benchmarking of different surfactant classes over these last years. Motivated by this limitation, we designed a theoretical methodology implemented in custom fast screening software to assess the surfactant activity on model lipid membranes. Experimentally based QSAR (quantitative structure-activity relationship) prediction delivered a set of parameters underlying the Diptool software engine for high-throughput agent-membrane interactions analysis. We validated our software by comparing score energy profiles with Gibbs free energy from the Adaptive Biasing Force approach on octenidine and chlorhexidine, popular antimicrobials. Results from Diptool can reflect the molecule behavior in the lipid membrane and correctly predict free energy of translocation much faster than classic molecular dynamics. This opens a new venue for searching novel classes of detergents with sharp biologic activity.

QSAR modelling using combined simple competitive learning networks and RBF neural networks

The aim of this study was to propose a QSAR modelling approach based on the combination of simple competitive learning (SCL) networks with radial basis function (RBF) neural networks for predicting the biological activity of chemical compounds. The proposed QSAR method consisted of two phases. In the first phase, an SCL network was applied to determine the centres of an RBF neural network. In the second phase, the RBF neural network was used to predict the biological activity of various phenols and Rho kinase (ROCK) inhibitors. The predictive ability of the proposed QSAR models was evaluated and compared with other QSAR models using external validation. The results of this study showed that the proposed QSAR modelling approach leads to better performances than other models in predicting the biological activity of chemical compounds. This indicated the efficiency of simple competitive learning networks in determining the centres of RBF neural networks.

Structure-activity relationship for Fe(III)-salen-like complexes as potent anticancer agents

Quantitative structure activity relationship (QSAR) for the anticancer activity of Fe(III)-salen and salen-like complexes was studied. The methods of density function theory (B3LYP/LANL2DZ) were used to optimize the structures. A pool of descriptors was calculated: 1497 theoretical descriptors and quantum-chemical parameters, shielding NMR, and electronic descriptors. The study of structure and activity relationship was performed with multiple linear regression (MLR) and artificial neural network (ANN). In nonlinear method, the adaptive neuro-fuzzy inference system (ANFIS) was applied in order to choose the most effective descriptors. The ANN-ANFIS model with high statistical significance (R²_train = 0.99, RMSE = 0.138, and Q²_LOO = 0.82) has better capability to predict the anticancer activity of the new compounds series of this family. Based on this study, anticancer activity of this compound is mainly dependent on the geometrical parameters, position, and the nature of the substituent of salen ligand.

A Quantitative Structure-Property Relationship Study on Photodegradation of Polybrominated Diphenyl Ethers

Quantitative structure-property relationship (QSPR) models were developed in the present work for photodegradation rate constants (kp) of fifteen individual polybrominated diphenyl ethers (PBDEs) in methanol/water (8:2) by UV light in the sunlight region. The molecular descriptors used in the QSPR models were calculated by the two semi-empirical quantum mechanical methods, RM1 and PM6, respectively. Both multiple linear regression (MLR) and artificialneural network (ANN) were applied in this study. The statistic qualities of the MLR models based on the molecular parameters obtained by RM1 and PM6 calculations were both good with the R values of 0.987 and 0.990, respectively. The QSPR model built by the ANN method with the molecular parameters calculated with PM6 is slightly better than that with RM1.