HMD-ARG: hierarchical multi-task deep learning for annotating antibiotic resistance genes

• DOI: 10.1186/s40168-021-01002-3
• Publication Date: 2021-02-08T22:50:31Z
• Authors: Yu Li; Zeling Xu; Wenkai Han; Huiluo Cao; Ramzan Umarov; Aixin Yan; Ming Fan; Huan Chen; Carlos M. Duarte; Lihua Li; Pak-Leung Ho; Xin Gao
• Keywords: Robustness; Sequence (biology)

Abstract Background The spread of antibiotic resistance has become one the most urgent threats to global health, which is estimated cause 700,000 deaths each year globally. Its surrogates, genes (ARGs), are highly transmittable between food, water, animal, and human mitigate efficacy antibiotics. Accurately identifying ARGs thus an indispensable step understanding ecology, transmission environmental human-associated reservoirs. Unfortunately, previous computational methods for mostly based on sequence alignment, cannot identify novel ARGs, their applications limited by currently incomplete knowledge about ARGs. Results Here, we propose end-to-end Hierarchical Multi-task Deep learning framework ARG annotation (HMD-ARG). Taking raw encoding as input, HMD-ARG can identify, without querying against existing databases, multiple properties simultaneously, including if input protein ARG, so, what family it resistant to, mechanism takes, intrinsic or acquired one. In addition, predicted beta-lactamase, further predicts subclass beta-lactamase that to. Comprehensive experiments, cross-fold validation, third-party dataset validation in gut microbiota, wet-experimental functional structural investigation conserved sites, demonstrate not only superior performance our method over state-of-art methods, but also effectiveness robustness proposed method. Conclusions We a hierarchical multi-task method, HMD-ARG, deep provide detailed annotations from three important aspects: class, mechanism, gene mobility. believe serve powerful tool and, therefore threat. Our constructed database available at http://www.cbrc.kaust.edu.sa/HMDARG/ .