BackgroundThe avian influenza A H5N1 virus occasionally infects humans, with high mortality rates. Although all current human infections are from avian-to-human transmission, it has been shown that H5N1 can be evolved to transmit between mammals, and is therefore a pandemic threat. For H5N1 surveillance, it is of interest to identify the avian isolates most likely to infect humans. In this study, we develop a method to identify mutations significantly associated with avian to human transmission.MethodUsing protein sequences for the surface glycoprotein hemagglutinin from avian and human H5N1 isolates in China, Egypt, and Indonesia from the years 1996–2011, we used Principle Component Analysis and a Maximum Likelihood Multinomial method to identify mutations associated with avian to human transmission. In each geographic region, transmission bias residues were identified using two signatures: a) significantly different amino-acid frequencies in human isolates compared to avian isolates from the same year, and b) significantly low probability of neutral evolution of the human isolates from the avian viral pool of the previous year.ResultsIn each geographic region, we find specific transmission bias mutations associated with human infections. These mutations are located in antigenic regions and receptor binding, glycosylation and polybasic cleavage sites of HA. We show that human isolates derive from a limited, subset of the avian pool characterized by geography specific mutations. In Egypt, two of three PCA clusters have very few human isolates but are highly enriched in mutations associated with a vaccine escape mutant H5N1 avian sub-clade that is known to be resistant to the Mexican H5N2 vaccine Furthermore, at these transmission bias associated residues, the mutations characteristic of these two clusters are distinct from those associated with the cluster enriched in human isolates, suggesting that vaccine resistant avian strains are unable to infect humans. Our results are relevant for surveillance and vaccination strategies for human H5N1 infections.