Antibiotic resistance has become one of the most urgent threats to global health, as more drugs are losing sensitivity to bacterium they were designed to kill. When investigating it, researchers usually need to identify and annotate antibiotic resistance genes (ARGs) from environmental or clinical samples.

Despite the existence of several computational tools for performing ARG annotation, most of them rely on sequence alignment, which can result in false negatives and biased prediction because of the incompleteness of the databases. In addition, most existing computational tools provide no information about the mobility of genes and the underlying mechanisms of resistance. 

To address such limitations, we propose an end-to-end Hierarchical Multi-task Deep learning framework for Antibiotic Resistance Gene annotation(HMD-ARG), taking raw sequence encoding as input and then annotating ARGs sequences from three aspects: resistant drug type, the underlying mechanism of resistance, and gene mobility. HMD-ARG can potentially serve as a tool for detecting antibiotic resistance with high accuracy.