Finding Protein-Binding Nucleic Acid Sequences Using a Long Short-Term Memory Neural Network

Abstract

With an increasing amount of data of protein-nucleic acid interactions, several machine learning-based methods have been developed to predict protein-nucleic acid interactions. However, most of these methods are classification models either for finding binding sites within a sequence or for determining whether a pair of sequences interacts. In this paper we propose a generative model for constructing nucleic acids binding to a target protein using a long short-term memory (LSTM) neural network. Nucleic acid sequences generated by the model showed high affinity for several target proteins. The generative model will be useful for constructing an initial library of nucleic acid sequences for in vitro selection of nucleic acid sequences that bind to a target protein with high affinity and specificity.