Three applications of neurocomputing in biomedical research

Summary form only given, as follows. Three different applications of neural networks in biomedical research are given. The frequency of side effects of 17 different drugs was predicted with less than 5% error using four physicochemical properties of the drugs. The six-month outcome of patients with a head injury was predicted with 86% of cases correct using five features of the patient determined at admission. Five different toxins were administered to rats and the toxin was correctly identified in 98% of the cases using the plasma bile acid profile. Backpropagation of errors was used for all three results. It is concluded that neural networks can be used for drug design and discovery and to provide information for patient care.<>