QSAR QSTR OF FLUOROQUINOLONES - AN EXAMPLE OF SIMULTANEOUS ANALYSIS OF MULTIPLE BIOLOGICAL-ACTIVITIES USING NEURAL-NETWORK METHOD/

In this paper, a three-layer three-output-node back-propagation neural network was applied to analyze the QSAR and QSTR of 23 5,8-disubstitu ted fluoroquinolones simultaneously. Four descriptors, q(N1), q(O4), s igma(m5), and MR8, were selected from a set of quantum chemical indice s and physicochemical parameters using PLS method. The results obtaine d from neural networks showed that the 4 descriptors were correlated s ignificantly with both the antibacterial activity and cytotoxicity and a multiple QSAR model was built. It provided useful information to de velop new selectively potent fluoroquinolones with high activity and l ow toxicity. Meanwhile, the study demonstrated that the multi-output-n ode neural network was a powerful tool to analyze multiple QSAR and su perior to PLS method. (C) Elsevier, Paris.