It would be important to measure the effect closely

Professor David Warhurst

This could block the flies' ability to transmit the malaria parasite, or make them particularly susceptible to pesticides.

The scientists say they would only have to release relatively few GM mosquitoes to kick-start the process.

According to the World Health Organisation, around 500m people a year are infected with malaria.

Any attempt to eradicate the mosquitoes that carry the disease as until now been doomed to failure because the flies are so numerous.

Infected flies

Dr Stephen Davis and his team at the Commonwealth Science and Industry Research Organisation in Canberra, Australia, think they have found a way to safely "infect" whole populations of mosquitoes with detrimental genes.

New Scientist magazine reports that their idea is to release engineered males that have two copies of a "type A" gene, and two copies of "type B".

Subsequently, individuals who inherit A and B together will be fine. But those who inherit either A without B or vice versa will die.

Computer models show that modifying just 3% of the population is enough to spread the genes.

Dr Davis said: "Such a low threshold was a bit of a shock to us."

Hybrid offspring

Hybrid offspring from matings between engineered males and wild females are fine because they have one A and one B.

But things get interesting when the hybrids start mating.

While all the offspring from hybrid/ engineered crosses are fine, some of the offspring from the hybrid/wild crosses die because they have an A or a B in isolation.

This creates a selection pressure which drives the genes through the population, as the offspring of wild flies die more often than those of engineered or hybrid flies.

This "drive mechanism" could be exploited to tackle malaria.

Other genes, that either kill the malaria parasite or make the flies susceptible to insecticides, can be tagged onto the modified genes, and will also spread through the population.

Professor David Warhurst, of the Public Health Laboratory Service malaria reference laboratory, said several groups were working on similar approaches.

He told BBC News Online: "This is apparently feasible, but it would be important to measure the effect closely, so to start with this would have to be introduced in a fairly selected area."