Quantitative Structure-Pharmacokinetic Relationships: Artificial Neural Network Modeling

Abstract

The technology and research revolution has provided many areas of science and industry with tools for more extensive and efficient operation. Nowhere is this phenomenon more evident than for new drug discovery and development in the pharmaceutical industry. Exploring the relationship between the structure of a molecule and its various biological and biochemical properties is the basis of drug discovery. Modern approaches to this field of study employ a combination of techniques. These include tests based on combinatorial chemistry and high-throughput (HT) screening as well as rational pharmaceutical design based on geometric and chemical characteristics of moleculemolecule interactions. Furthermore, understanding and optimising factors such as the effect of a compound on the body and the effect of the body on a compound are essential in developing a new drug. <br/> The main bottleneck in drug discovery is the identification of new chemical entities (NCEs) to be used for drug leads. The 1990s saw development of new automated tools for drug discovery including combinatorial chemistry and high-throughput screening. These tools have led to the increased discovery of new drug lead compounds each of which in tum require pharmacological and pharmacokinetic testing. Moreover, substantial increases in computing power as well as development of robust software has given scientists the opportunity to undertake significant research projects from their own desktops. Consequently, data analysis, data mining, and information manipulation have all benefited and progressed considerably. <br/> Software programs have been developed for a wide range of fields such as quantitative structureactivity relationship (QSAR) studies, pharmacophore elucidation, molecular modeling, drugreceptor interactions and in vivo simulations. Newer techniques have been influenced by what is termed "soft computing" which aims to accommodate the imprecision and uncertainty inherent in the real world [Zadeh, 1996]. Soft computing draws on the model of the human brain and derives mainly from artificial intelligence (Al) sources including genetic algorithm (GA), fuzzy logic, and artificial neural network (ANN) approaches [Maddalena, 1998]. Other less conunon techniques include cellular automata, fractals and chaos theory. ANNs are aparticularly useful modeling tools for nonlinear systems. Although not as common in the pharmaceutical industry as conventional modeling and mathematical techniques, soft computing has been successful in a number of fields in the industry.