Description:
Various methods have been proposed to search for molecules with the desired properties. However, it remains difficult due to the vastness of the search space. Because of the reason, it is not easy to obtain a suitable solution by applying randomized algorithms.In recent years, deep learning-based methods have achieved remarkable results in various fields. Since most of the methods rely on existing datasets, it is difficult to remove bias derived from the dataset. In addition, the annotation cost of preparing new data sets is also significant.This study offers some important insights into the scheme of generative models. We propose Directional Generative Networks (DGN), molecular candidates finding method using a regression model. The method can directly generate molecular candidates that satisfy the desired molecular features even in the absence of molecular data sets