Early detection of pathogens with pandemic potential is of clear public health importance. Spatial prediction models allow targeted surveillance of geographic ‘hotspots’ to monitor new emerging diseases. Importantly, however, the vast majority of emerging diseases have resulted in dead-end ‘spill-over’ events in which a pathogen is transmitted from an animal reservoir to a human, but is then unable to achieve the sustained human-to-human transmission that is necessary for disease pandemics. It is therefore critical to determine why, under similar ecological settings, some virus infections are transmitted among humans while others are not. We sought to determine whether there are specific biological features that characterize a ‘successful’ human pathogen; that is, those viruses capable of transmitting between humans. To this end we compiled a database of 116 human viruses and within them analyzed a set of variables including: taxonomic classification, genome length, type and structure; outer protein status; estimates of recombination frequency; as well as duration of infection and mortality. Using an information theoretic approach we assessed the power of these variables as predictors of human-to-human transmission, which revealed multiple significant associations. In particular, we revealed that all viruses that are able to establish chronic infections in humans were able to achieve human-to-human transmission. Focusing our statistical analysis to viruses with acute infection, we found that the structure of the virus, recombination rate and mortality rate were all strong predictors of human-to-human transmission, while genomic traits of the virus were of less importance. The model developed here can be used to generate predictions of the probability that a specific virus will achieve human-to-human transmission.