Researchers from Western University recently made a splash with news that a HIV vaccine (SAV001) has made its way through initial clinical trials, and will be now tested on humans.
“We were very excited with the Phase I results,” the researchers say. “The trial demonstrated that our vaccine stimulates broadly neutralizing antibodies that will neutralize not only single sub-types of HIV, but other sub-types, which means that you can have the vaccine cover many different strains of the virus.”
Despite this apparent progress however, there is still clearly much to be done to tackle a disease that affects some 17 million people around the world. This number remains high despite the number of people being infected by the HIV virus actually going down, with this apparent dichotomy largely because a cocktail of treatments is able to keep patients alive for longer.
Mixing the Cocktail
Finding the right treatment for each person remains a challenge however, and is largely am ad-hoc search by physicians to best treat the specific mutation that a particular patient comes with.
Researchers from IBM have developed a machine learning based approach to try and shorten this search process, whilst simultaneously predicting the success of the cocktail. What’s more, researchers believe that such methods can also be used to predict how long it will take for the virus to develop a form of immunity to the drugs.
The work is being done under the EU project, called EuResist, which began several years ago to work on optimizing treatments, including that of AIDS. The tool is offered freely throughout the EU, and taps into data from across both Europe and Africa to find the right combination of drugs to provide resistance for the maximum amount of time. Crucially, it can also provide doctors with a good idea of when the drugs may cease being effective.
Initial tests of the approach suggest that it is capable of outperforming human doctors when it comes to predicting the effectiveness of treatment, which lends further weight to previous studies highlighting the effectiveness of algorithms for these kind of tasks.
With current researchers looking to enable the algorithms to match the current mutation with the best treatment, this could provide doctors with a degree of adaptability that would have been difficult to match.