BERT,SOM,CNN を用いた DNA 配列分類手法

抄録

One of the great goals in biology is to unravel the DNA sequence, which is the blueprint of an organism. Recently, it has become possible to obtain large amounts of DNA sequence information with the development of DNA sequencers. Especially in the field of metagenomics, it is necessary to distinguish each species from DNA sequence fragments obtained from multiple species. For this purpose, analysis method of DNA sequence fragment information using the combination of DNABERT which is a method based on natural language processing, Self Organizing Map, and Convolutional Neural Network is proposed, and the performance is confirmed in some experiments.