NEURAL-NETWORK APPROACHES APPLIED TO SELECTIVE A(2A) ADENOSINE RECEPTOR AGONISTS

The generalized-regression genetic-neural network was tested using a s eries of selective A(2a) adenosine receptor agonists (2-aralkynyl and 2-heteroaralkynyl derivatives of adenosine-5'-N-ethyluronamides), It w as shown that the model allows to select physicochemical descriptors o f molecules, and it gives much more better results than a backpropagat ion neural network.