Prediction of Nucleosome Forming and Nucleosome Inhibiting DNA Sequences Using Convolutional Neural Networks

The nucleosome is the fundamental unit of DNA packaging in eukaryotic cells. Nucleosomes play a major role in the regulation of gene activity as they act like barriers that control the transcription by allowing or blocking DNA-binding proteins, transcription factors, and activators from accessing to DNA. In this paper, we proposed a deep learning method based on a convolutional neural network (CNN) that can accurately distinct nucleosome-forming from nucleosome-inhibiting DNA sequences. We utilize dataset of three different organisms, and for each dataset, we conduct two different experiments. We exploited $k$-mer frequency count as input as well as using the DNA sequence directly. The proposed CNN models nicely recognize the different DNA sequences by learning the key region and features that characterize each DNA sequence type. We compare our result with the previous works based on the evaluation conducted on three different datasets including Human, Fly, and Worm. The proposed CNN models achieved comparable results.