Research of a Recurrent Neural Network for the Vector Representation of Nucleotide Sequences

Embeddings can be a good way to work with gene sequences, helping to avoid the high computational complexity problems that are often found in multiple sequence alignment. In this paper, we investigate the possibilities of the LSTM network for constructing embeddings of nucleotide sequences. To do this, we trained model on a dataset of 1.5 million pairs of gene sequences belonging to the Escherichia coli species.