Hybrid 3-D QSAR approach using a generalized-regression genetic-neural network
Peter P. Mager · Molecular Simulation · 2002
The advantage of generalized-regression genetic-neural networks over backpropagation networks is that relevant physicochemical descriptors can be selected, somewhat reminiscent of the results from classical quantitative structure-activity relationships (QSARs). The descriptor matrix is based on a hybrid model of traditional QSAR (use of extra-thermodynamic parameters) and 3-D QSAR. The model was tested using molecular simulation of a series of A 3 adenosine receptor agonists (1,4-dihydropyridine derivatives).