Previous PageNext Page

1. Introduction

The development of a new drug is an extremely laborious and time-consuming process. Thus, quite early on, computer methods have been used to further an understanding of the interactions of a drug with its receptor. Molecular modelling and rational drug design have become indispensable tools for the development of a new drug [1]. Recently, combinatorial chemistry and high-throughput screening have been introduced in order to speed up the drug development process. These methods produce massive amounts of data that have to be analyzed in an efficient manner in order to make best use of these novel methods. We will show here that self-organizing neural networks, such as the one introduced by Kohonen [2], can be used both in rational drug design and in combinatorial chemistry.
The application of neural networks in chemistry has increased dramatically in recent years [3]-[5]. In a Kohonen neural network (KNN), the artificial neurons self-organize in an unsupervised learning process and, thus, can be used to generate topological feature maps. It will be shown here that this potential can be utilized to analyze the shape and surface properties of those three-dimensional objects responsible for biological activity, molecules.
In these applications, there is a one-to-one mapping of a single molecule into a single Kohonen network. However, a Kohonen network can also be used for the analysis of datasets of molecules, where several molecules are simultaneously mapped into one Kohonen network. In order to make full use of the potential of self-organizing networks, novel representations of molecular structures have been developed. These methods can be put into a clear hierarchy, starting from molecular topology going all the way to molecular surfaces. They do not only encode structural information, but also information on the properties of atoms or of molecular surfaces.

Previous PageNext Page


Johann.Gasteiger@chemie.uni-erlangen.de