Deep neural networks are a set of computer algorithms within machine learning that, by layering a large number of artificial neurons, allow the approximation of complex multidimensional functions. A prerequisite for the implementation of each machine learning algorithm is the definition of a model, an empirical error and an optimization procedure that minimizes the empirical error. Also, due to the very high degree of expressiveness of deep neural networks, we need a sufficiently large and high-quality input data set. This paper describes the functionality of some of the more important groups of deep neural networks such as convolutional and recurrent neural networks. Deep neural networks have been applied to a wide range of problems and have thus found application in the field of drug development and repurposing. Characteristics of a molecule such as the ability to react with other molecules and ADMET properties can be predicted from the representation of the molecule itself such as the SMILES representation. Also, a set of generative neural networks has been developed that can produce new and unseen specimens of representations of molecules that have similar properties as the molecules of the input data set. The use of deep neural networks for drug repurposing for other indications allows expansion of the existing set of indications of a drug that has already been in clinical use and for which the properties and side effects profile are known, thus reducing the time required to be approved for other indications. All of the above leads to a more efficient process of drug development, which benefits all participants in the chain of production and distribution of drugs, from the pharmaceutical companies to patients.