Abstract
Quantitative structure–activity relationship (QSAR) models were derived from a structurally heterogeneous set of 200 phenol derivatives for which the 50% growth inhibition concentration (IGC50) values to the ciliated protozoan Tetrahymena pyriformis were available. Each molecule was described by means of physicochemical descriptors and structural features. Partial least squares (PLS) regression analysis and a three-layer perceptron were used as statistical engine. The performances of the linear and nonlinear models were estimated from an external testing set of 50 chemicals. Despite hard constraints voluntarily imposed in the design of the neural network models, they provided better simulation results than the PLS models.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have