Applying CNN to Cheminformatics

We describe a method for the construction of specific types of neural networks composed of structures directly linked to the structure of the molecule under consideration. Each molecule can be represented by a unique neural connectivity problem (graph) which can be programmed onto a cellular neural network. A composite network can further successfully perform classification and regression on real-world chemical data sets. The method can be regarded as a statistical learning procedure that turns the molecular graph, representing the 2D formula of the compound, into an adaptive whole molecule composite descriptor. By translating the molecular graph structure into a dynamical system, the algorithm can compute an output value that is highly sensitive to the molecular topology. This system can be trained by gradient descent techniques which rely on the efficient calculation of the gradient by backpropagation.