Bacteria Taxonomic Classification using Graph Neural Networks

Abstract

Graph neural networks are effective and useful tools for problems that may be represented using graphs. The De Bruijn graph is a directed graph used to express overlaps between sequences of symbols in DNA sequence representation. In this paper, we present a method for sequence categorization using a De Bruijn sequence representation and a Convolutional Graph Neural Network (GCNN). We tested the methodology on a classification problem involving the 16S gene sequences. An analysis conducted on a dataset of 3000 16S sequences demonstrates results in comparison to state-of-the-art. The dataset utilized in the research and the source code can be accessed at https://github.com/Calder10/Bacteria-Taxonomic-Classification-using-GNN.