Design of New Biologically Active Molecules by Recursive Neural Networks

摘要

In this paper, we face the design of novel molecules belonging tothe class of adenine analogues (8-azaadenine derivates), that present awidespread potential therapeutic interest, in the new perspectiveoffered by recursive neural networks for quantitative structure-activityrelationships analysis. The generality and flexibility of the methodused to process structured domains allows us to propose new solutions tothe representation problem of this set of compounds and to obtain goodprediction results, as it has been proved by the comparison with thevalues obtained “a posteriori” after synthesis andbiological essays of designed molecules