A Hierarchical Classification Model for Annotating Antibacterial Biocide and Metal Resistance Genes via Fusing Global and Local Semantics

Xing Lv, Jie Deng, Weizhong Zhao, Xinhui Tu, Xingpeng Jiang

Abstract: Bacteria with resisting antibacterial biocide and metal agents have become a commonly global concern for public health and environment protection. Since the resistance of bacteria is mainly attributed to expression of certain genes, accurately annotating the antibacterial biocide and metal resistance genes is the fundamental step to address or mitigate the global concern. However, due to the complex biological mechanisms of bacterial resistance for antibacterial biocides and metals, existing prediction methods face challenges in accurately categorizing antibacterial biocide and metal resistance genes, leading to undesirable prediction performance. In this paper, we propose a Hierarchical Classification Model for annotating Antibacterial Biocide and Metal Resistance Genes (HCM-ABMRGs) by considering the global and local semantics contained in gene-coded protein sequences. More specifically, the task of annotating antibacterial biocide and metal resistance genes is treated as a three-level classification problem, in which properties of genes are annotated from coarse to refined granularity. In addition, global and local semantics contained in protein sequences are explicitly captured to derive more meaningful representations for annotating genes. Comprehensive experiments are conducted on a widely used dataset, and experimental results demonstrate the effectiveness of the proposed method.