The emergence and rapid dissemination of antibiotic resistance threatens medical progress and calls for innovative approaches for the management of multidrug resistant infections. Phage-therapy, i.e., the use of viruses that specifically infect and kill bacteria during their life cycle, is a re-emerging and promising alternative to solve this problem. The success of phage therapy mainly relies on the exact matching between both the target pathogenic bacteria and the therapeutic phage. Several papers propose models for in-silico prediction of phage-bacteria interactions but at the species level. In clinical applications prediction of phage-bacteria interaction at species level is not enough to target a given pathogenic bacteria strain.One of the main challenges to train classification models able to predict phage-bacteria interactions is the need of both types of samples: interaction and non-interaction phage-bacteria couples. Non-interactions are rarely reported, making these data scarce. This problem is even more evident for non-interaction data at strain level. These factors make difficult the use of classic machine learning algorithms which need relatively-balanced classes to produce accurate predictions.This problem calls for solutions to deal with such imbalanced data. In this paper are presented two approaches to tackle this problem. 1. To explore the use of One-Class learning methods 2. To generate putative non-interacting data and use single and ensemble-learning approaches, to predict phage-bacteria interaction at strain-level.