Artificial Intelligence Applications in Bacteriophage Genomics and Host Interaction

Antibiotic-resistant bacterial infections remain a major public health challenge, contributing to an estimated 1.27 million deaths in 2019. Pathogens, such as Escherichia coli, Klebsiella pneumoniae, and Acinetobacter baumannii, cause a range of infections in clinical and food environments. As these bacteria rapidly develop resistance to multiple antibiotics, there is a growing need for alternative treatments. One promising approach is bacteriophage (phage) therapy. Phages are bacterial viruses that recognize, infect, and kill their hosts to propagate. Their relatively low isolation and production costs make them attractive antibacterial agents. However, unlike antibiotics, their host specificity requires individualized testing to identify effective phages, typically through plaque assays. This labor-intensive and time-consuming method limits the scalability of phage therapy, especially for large phage libraries. With the remarkable advances in genome sequence technologies over the past years, a vast and highly diverse set of phage genomes have been sequenced at decreasing costs. This growing volume of genomic data has enabled the application of Artificial Intelligence (AI), particularly Machine Learning (ML), and its subfield Deep Learning (DL) in phage research. Data-driven approaches are increasingly used for genome annotation, host prediction, lifestyle classification, and the identification of therapeutic candidates, offering scalable alternatives to traditional methods. Despite the availability of numerous genome annotation tools, only a limited number are specifically tailored to phage genomes. Phage infectivity, determined by host range, arises from complex interactions between the phage and its host at both extracellular and intracellular levels. Extracellularly, successful infection depends on the phage's ability to recognize and bind to host receptors. Intracellularly, bacteria have evolved a variety of innate and adaptive defense systems to prevent phage replication. To better understand these interactions, several computational tools have been developed to detect and characterize phage-host interactions. A comprehensive understanding of phage-host dynamics, both extracellular and intracellular level, is critical for advancing predictive models and facilitating the clinical translation of phage therapy.