QUANTITATIVE STRUCTURE-ACTIVITY-RELATIONSHIPS BY NEURAL NETWORKS AND INDUCTIVE LOGIC PROGRAMMING .1. THE INHIBITION OF DIHYDROFOLATE-REDUCTASE BY PYRIMIDINES

Neural networks and inductive logic programming (ILP) have been compar ed to linear regression for modelling the QSAR of the inhibition of E. coli dihydrofolate reductase (DHFR) by 2,4-diamino-5-(substituted ben zyl)pyrimidines, and, in the subsequent paper [Hirst, J.D., King, R.D. and Sternberg, M.J.E., J. Comput.-Aided Mel. Design, 8 (1994) 421], t he inhibition of rodent DHFR by ,4-diamino-6,6-dimethyl-5-phenyl-dihyd rotriazines. Cross-validation trials provide a statistically rigorous assessment of the predictive capabilities of the methods, with trainin g and testing data selected randomly and all the methods developed usi ng identical training data. For the ILP analysis, molecules are repres ented by attributes other than Hansch parameters. Neural networks and ILP perform better than linear regression using the attribute represen tation, but the difference is not statistically significant. The major benefit from the ILP analysis is the formulation of understandable ru les relating the activity of the inhibitors to their chemical structur e.