No abstract

The accuracy of in silico models can be inhomogeneous: models can show excellent performance on some chemical subspaces but have low accuracy on others. We show that applicability domain (AD) approaches can differentiate reliable and non-reliable predictions and identify those with experimental accuracy for both regression and classification models. For reliably predicted molecules, the predicted values can be used instead of experimental measurements. This can halve time and costs of experimental measurements. The developed classification models for AMES mutagenicity test and CYP450 inhibition, which are important drug discovery properties, are publicly available at the online chemical database and modeling environment (OCHEM) site http://qspr.eu

Azetidin-2-one analogues are reported to exhibit various pharmacological activities like cholesterol absorption inhibitory activity, human tryptase, thrombin and chymase inhibitory activity, vasopressin V1a antagonist activity, antidiabetic, anti-inflammatory, antiparkinsonian and anti-HIV activity in addition to antimicrobial. 1-6 In the present study, Isoniazid (INH), the established antitubercular drug was selected as the lead for the design and development of antitubercular agents with minimal toxic effects. A novel series of amino azetidinones were designed from corresponding azetidin-2-ones using various in silico methods. Docking studies were performed at Mtb enoyl acp reductase (4DRE) and the derivatives exhibited best docking scores were prepared from corresponding azetidin-2-ones by treating with various molecules containing amino groups in the presence of TEA. Azetidin-2-ones in turn were obtained from a series of INH Schiff bases by reaction with chloro acetyl chloride. Structures of the newly synthesized compounds were assigned on the basis of elemental analysis, IR, 1 H NMR, 13 CNMR and mass spectral studies. The newly synthesized compounds were screened for their in vitro antitubercular activity using Alamar blue assay method and the hepatotoxicity was determined by MTT assay method using chang liver cells. AAZ1V, the amino azetidinone obtained from N-[3-chloro-2-(4-chlorophenyl)-4-oxoazetidin-1-yl] pyridine-4-carboxamide (AZ1V) by combining with 4-amino 1, 2, 4-triazole produced significant antitubercular activity. The percentage viability produced by AAZ1V against Chang liver cells for hepatotoxicity was better than the percentage viability produced by INH.

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 structureactivity 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. The use of variables selected by the elaborated methods provides an improvement of neural network prediction ability compared to that calculated using the unpruned sets of variables.