Many are worried H5N1 will mutate and pass from human-to-human
International scientists believe they have identified some of the key steps needed for bird flu to develop into the deadly pandemic strain of the disease.
The team pinpoints two genetic mutations that would need to occur to the H5N1 virus for it to potentially spread readily between humans.
Writing in the journal Nature, the scientists said the findings would help them to detect pandemic strains.
At present, H5N1 can pass only from bird-to-bird or, rarely, bird-to-human.
So far, there have been a total of 258 cases of H5N1 in humans, causing 153 deaths, according to figures from the World Health Organization.
But flu viruses mutate and evolve quickly, and scientists believe the virus could acquire the ability to pass between humans.
They fear this could trigger a repeat of the devastation of the 1918 flu pandemic, which is thought to have killed 50 million people.
To investigate how the virus might do this, the researchers looked at samples of H5N1 that had been taken from birds and also from infected humans.
In a small number of the human samples, they found the virus had acquired small changes to a protein called haemaggluttinin, which sits on the surface of the H5N1 molecule.
This protein helps the flu virus to spread by binding to the receptors on cells, which are like docking stations, allowing the virus to invade and infect the cells.
While the haemaggluttinin in most of the samples could only bind to bird cell-receptors, the researchers discovered that in some of the human samples, the haemaggluttinin had acquired the ability to bind to both bird and human cell-receptors.
It is thought this is a key step needed for H5N1 to be able to spread from human to human.
Further analysis revealed two separate mutations at different positions on the protein had enabled H5N1 to recognise human receptors.
The researchers said the discovery of the location of the mutations would help identify H5N1 strains that might be on the way to developing pandemic potential.
Lead researcher Yoshihiro Kawaoka, a virologist at the University of Wisconsin-Madison School of Veterinary Medicine, US, said more mutations would be required for the virus to fully adapt to humans, but it is not known how many mutations are needed for such a change.
The team thought these changes were most likely to occur when a human influenza virus mingled with H5N1, particularly if it could already bind to human receptors.
Dr Wendy Barclay, a molecular virologist from Reading University, UK, said: "This work shows that at least two changes in the haemaggluttinin protein are needed for H5N1 to transform to strain that could infect humans, and knowing what these are will help to inform surveillance."
But, she said, the fact these mutations had already been seen in viruses isolated from human H5N1 cases, and a pandemic had not yet struck, suggested a number of other steps might be needed for the virus to be able to pass from human to human.