COMPUTER-AIDED DRUG DESIGN - A NEURAL-NETWORK APPROACH

Artificial neural networks have received considerable attention in recent years as their particular properties have come to be recognised, especially their ability to learn and tolerance to noise. In this paper, we describe how some of the properties of neural networks can be used to assist the medicinal chemist in the design of drugs. Specifically, the back propagation neural network has been used to determine the partition coefficient and aqueous solubility of a number of organic compounds. The partition coefficient and aqueous solubility were chosen since they are known to correlate well with biological activity. The results obtained show that the neural network after training can be used to determine the partition coefficient and solubility of a series of unknown compounds to an accuracy comparable to the experimental error. Using the partition coefficient data we have conducted an extended series of experiments into the effect that network parameters have on performance. The results show that performance of the neural network can be significantly affected by the choice of network parameters. In conclusion, we propose a system for predicting pharmacological activity incorporating an expert system and multiple neural networks and describe the benefits that such a system may offer.