No abstract

The aim of this study was to develop a simple quantitative structure-activity relationship (QSAR) for the classification and prediction of antibacterial activity, so as to enable in silico screening. To this end a database of 661 compounds, classified according to whether they had antibacterial activity, and for which a total of 167 physicochemical and structural descriptors were calculated, was analyzed. To identify descriptors that allowed separation of the two classes (i.e. those compounds with and without antibacterial activity), analysis of variance was utilized and models were developed using linear discriminant and binary logistic regression analyses. Model predictivity was assessed and validated by the random removal of 30% of the compounds to form a test set, for which predictions were made from the model. The results of the analyses indicated that six descriptors, accounting for hydrophobicity and inter- and intramolecular hydrogen bonding, provided excellent separation of the data. Logistic regression analysis was shown to model the data slightly more accurately than discriminant analysis.

Metabolic drug-drug interactions are receiving more and more attention from the in silico community. Early prediction of such interactions would not only improve drug safety but also contribute to make drug design more predictable and rational. The aim of this study was to build a simple and interpretable model for the determination of the P450 enzyme predominantly responsible for a drug's metabolism. The P450 enzymes taken into consideration were CYP3A4, CYP2D6 and CYP2C9. Physico-chemical descriptors and structural descriptors for 96 currently marketed drugs were submitted to statistical analysis using the formal inference-based recursive modelling (FIRM) method, a form of recursive partitioning. Generally accepted knowledge on metabolism by these enzymes was also used to construct a hierarchical decision tree. Robust methods of variable selection using recursive partitioning were utilised. The descriptive ability of the resulting hierarchical model is very satisfactory, with 94% of the compounds correctly classified.

Quantitative structure-activity relationship (QSAR) analysis of four toxicological data sets is described. The toxicological data include three data sets retrieved from the literature (the toxic and metabolic effects of 23 aliphatic alcohols on the perfused rat liver; the toxicity of 21 pyridines to mice; the lethality of 55 halogenated hydrocarbons to the mould Aspergillus nidulans). In addition, the toxicity of 13 mono- and di-substituted nitrobenzenes in a 15 min assay using the alga Chlorella vulgaris was analysed. QSARs were developed successfully using descriptors to describe uptake in the organism (i.e. hydrophobicity as quantified by the logarithm of the octanol-water partition coefficient, log P) and reactivity at the site of action (i.e. electrophilicity as quantified by the energy of the lowest unoccupied molecular orbital, E(LUMO)). A further parameter describing molecular branching as also required to model the data for the aliphatic alcohols. The results demonstrate that mechanistically based QSARs can be developed for these diverse endpoints which are, in terms of statistical quality as good as, if not better, than QSARs based on less mechanistically interpretable descriptors.

Three methods for estimation of the equilibrium tissue-to-plasma partition ratios (Kp values) in the presence of tissue concentration time data have been investigated. These are the area method, the open loop (tissue specific) method and the whole body model(closed loop) method, each with different model assumptions. Additionally, multiple imputations, a technique for dealing with deficiencies in data sets (i.e., missing tissues) is used. The estimated Kp values by the three methods have been compared and the limitations and advantages of each approach drawn. The area method, which is essentially model free, gives only a crude estimate of Kp without making any statement of its uncertainty; whereas both the open and closed loop methods provide an estimate of this. The closed loop method, where the most assumptions are made, is the approach that gives the best overall estimates of Kp, which was confirmed by comparing the predicted concentration-time profiles with experimental data. Although the estimates from the closed loop method, as well as the other two methods, are conditioned on the data, they are the most reliable for both propagating parameter variability and uncertainty through a whole body physiologically based model, as well as for extrapolation to human. A series of benzodiazepines, namely alprazolam, chlordiazepoxide, clobazam, diazepam, flunitrazepam, midazolam and triazolam in rat is used as a case study in the current investigation.