摘要：
      抗生素耐药性基因（ARGs）是全球细菌感染数量不断增加的原因。ARGs在细菌基因组中很难追踪，因为它们经常通过移动遗传元件（MGE）进行水平基因转移。复杂的蛋白质-蛋白质网络可以揭示有助于传播和持久性的蛋白质。在这里，我们开发了一个管道，通过分析蛋白质-蛋白质相互作用网络（PPIN）的特征来促进这一过程。该管道使用随机森林模型来区分ARGs和非ARGs，并探索ARGs与它们功能性相互作用的蛋白质之间的关联。我们使用大肠杆菌和鲍曼不动杆菌的PPIN测试了该方法，这两种已知携带MGE携带ARG的致命生物体，在ARG鉴定中实现了85%的宏观平均准确率。该方法还表明，ARGs与MGE的相关性不成比例，并且邻居（仅与ARGs的一个边缘连接的基因）的流动性可能较低。
Abstract:
Antibiotic resistance genes (ARGs) are responsible for an increasing number of bacterial infections worldwide. ARGs are challenging to track within bacterial genomes as they are often subject to horizontal gene transfer via mobile genetic elements (MGEs). Complex protein-protein networks can reveal proteins contributing to the spread and persistence. Here, we developed a pipeline that facilitates this process by analyzing features of a Protein-Protein Interaction Network (PPIN). This pipeline uses a random forest model to distinguish ARGs from non-ARGs and explores associations between ARGs and proteins with which they functionally interact. We tested the approach using the PPINs of Escherichia coli and Acinetobacter baumannii, two deadly organisms known to carry ARGs harbored by MGEs, and achieved a macro average accuracy of 85% in ARG identification. The approach also revealed that ARGs are disproportionately associated with MGEs and the neighbors (genes connected with only one edge to the ARGs) are likely to be less mobile.

https://ieeexplore.ieee.org/abstract/document/9995224