There is an urgent need for replacements for antibiotics. During evolution bacterial viruses have evolved to manipulate or destroy specific human pathogens. There are huge and rapidly increasing numbers of sequences from bacterial viruses in the public domain, but this wealth of data has not been able to be fully utilised. We believe that it has now become possible to mine this data for new antibiotics. In preliminary studies we have identified thousands of potential anti-bacterials – the difficulty is choosing the best ones. To select these we propose to develop novel computer methods to predict their hosts, biophysical, functional, and expression properties. If successful this research will provide proof of principle by discovering and testing pre-clinically new anti-bacterials. It will open up a novel and flexible computational approach to transform antibiotic discovery.