Since Edward Jenner's original experiments with cow-pox, researchers have sought new methods of vaccination that were safer, cheaper, and provided greater protection. The application of genomics technologies to this problem, called Reverse Vaccinology (RV), has the potential to revolutionize the identification of bacterial protective antigens (BPAs) for the development of new vaccines. In this study, we improved upon previously published RV methods by applying modern machine learning tools. We used Vaxijen, a previously published RV data set, in addition to curating our own data set of bacterial protective antigens from the primary literature. We generated training data by applying 18 publicly available programs for protein annotation to both data sets, and classifiers were then constructed using Support Vector Machines (SVMs), Partial Least-Squares Discriminant Analysis (PLS-DA), and Linear Regression. Accuracy was assessed with stratified leavetenth- out cross validation (LTOCV), for which we report a maximum accuracy of 92% with the BPA data set. This classifier was then assessed for its ability to recall of BPAs from the proteomes of known pathogens, for which we report p-values of 0.00664 (S. aureus) to <0.0001 (M. tuberculosis). To our knowledge this is the first time that machine learning has been used in Reverse Vaccinology, and our results suggest our methods would enhance current efforts to identify of protective antigens in silico.