|
|
|
|
|
|
|
curonyltransferase occur with compounds with a log P of 2.0 [10]an optimal value found for many distribution and binding processes in the body [11]. Similarly, using principal-component analysis, Holmes et al. [12] were able to predict the extent of sulphate and glucuronide conjugation of a series of substituted phenols. |
|
|
|
|
|
|
|
|
Recently certain computer software has become available to investigate these types of relationships using new algorithms. One such programme is TSAR (Tools for Structure Activity Relationships) from Oxford Molecular [13]. Using a structural-related molecular spreadsheet, it can not only calculate specific properties where necessary (charge properties; dipole moments, lipophilicity, Verloop indices; topological descriptors; Kier analysis, flexibility indices, and shape recognition, etc.), but also relate them to any measure, be it pharmacology, toxicology, or, for this discussion, kinetic and metabolic rates and pathways. Using a training set of data from a certain molecular series, univariate and multivariate analysis, such as principal-component analysis, discriminate analysis, hierarchical classification clusters, and neural networks [14] (see later), it is possible to predict the major endpoints of biodisposition for new compounds based on information from a few key structures. Others have used similar approaches in conjunction with fuzzy logic to obtain kinetic parameters from QSAR with some success [15]. |
|
|
|
|
|
|
|
|
A similar approach has been suggested [16] for the prediction of toxicity caused by the reactions catalyzed by the P450 cytochromes using a Computer-Optimized Molecular Parametric Analysis of Chemical Toxicity (COMPACT). Similarly, it is envisaged that with the increasing research on the structural interactions of drugs with cloned and amino acid sequenced cytochromes and other metabolic enzymes [17], QSAR will in the future be used to rapidly screen the thousands of drugs produced by combinatorial chemistry and predict which enzyme will metabolize which drug. |
|
|
|
|
|
|
|
|
2. Data-Based Expert Systems |
|
|
|
|
|
|
|
|
The approaches outlined above should be able to provide a guesstimate of certain kinetic parameters, providing that, in any one series, some data are already available for a few compounds. However, the prediction of which metabolic pathway is involved and what products will be formed will continue to be a problem. To overcome this, several expert systems and interrelational databases have been developed. Early systems, such as Xeno [18], have been reported but failed to be of general usefulness. Similar difficulties have been encountered for Metabol Expert [19] and Meta 2 [20], which give many possible metabolic pathways but provide little assurance as to whether the metabolites suggested actually can be expected when compared to real data [21]. These systems fail since they are dependent on the actual information in the |
|
|
|
|
|