PREDICTION OF MATERIAL PROPERTIES FROM CHEMICAL STRUCTURES - THE CLEARING TEMPERATURE OF NEMATIC LIQUID-CRYSTALS DERIVED FROM THEIR CHEMICAL STRUCTURES BY ARTIFICIAL NEURAL NETWORKS

The prediction of properties of molecules just on the basis of their c hemical structures is desirable to selectively make molecules that hav e the wanted properties, like biological activity, viscosity, or toxic ity. Here, we present an example of a new way to predict a property fr om the chemical structure of a chemically heterogeneous class of compo unds. The clearing temperatures of nematic liquid-crystalline phases o f 17 383 compounds were used to train neural networks to derive this m aterial property directly from their chemical structure. The trained n eural networks were subsequently tested with 4345 structural patterns of molecules unknown to the networks to assess their predictive value. The clearing temperatures were predicted by the best network with a s tandard deviation of 13 degrees.