Given the problems with antimicrobial resistance, bacteriophages are being investigated as promising candidates for treating multidrug resistant bacteria. Because phages’ range of infection is usually at the strain level, selecting the appropriate phage for a given bacterial infection is challenging. In this project, we implement machine learning (ML) analyses to understand the biology underlying phage-bacteria interactions and to make better decisions when selecting phages for therapy. We worked with a library of 314 Uropathogenic Escherichia coli strains challenged against 31 phages. The bacterial growth kinetics were measured in liquid assays. Additionally, the bacteria and phages were whole genome sequenced, and pangenome analyses were performed [1]. The performance of phages infecting bacteria was measured in terms of the area under the curve, height of the peak, survival proportion, and regrowth rate. Correlations between these metrics demonstrate they possess complementary information, as previously shown in silico [2]. In addition, we trained ML models to predict the values of these quantities for new bacterium-phage pairs, based on their genomes. In our pipeline, a recursive feature elimination process is done, from which we can obtain the most statistically relevant genes, both in bacteria and phages. This can give valuable insights into the biological processes involved, which is crucial for engineering new phages. In the future, we will implement ML algorithms to classify the interactions based on the different metrics. We anticipate that the resulting clusters will provide more meaningful guidance than single measurements when selecting phages for therapies. [1] doi.org/10.1073/pnas.2313574121 [2] doi.org/10.1073/pnas.2513377122