In one of the first demonstrations that personalised medicine may be achievable in the near future, researchers from University College London, UK have shown that potential HIV drugs can be screened against a virtual model of a patient in short enough time scales to be of clinical use.
The researchers used grid computing services from the UK and the USA to perform the simulations in less than two weeks. The research, published in the Journal of the American Chemical Society, makes use of the new Virtual Physiological Human – an EU-funded project that hopes to recreate the inner workings of the human body in silico.
Personalised therapies will be particularly useful for HIV treatment as the virus mutates very quickly within the patient, with different drugs affecting each mutation differently. The way the virus mutates can depend on a number of factors, including the patient’s genetic makeup, previous treatment and other environmental factors, so it is not normally possible for doctors to predict which treatments would prove effective.
Previously, doctors would need to test each drug separately on the patient to see which ones worked best. It is hoped that the new simulations would replace this trial and error process to provide the right drugs from the very start.
The models are built in two distinct stages. Firstly, the potential mutations of the virus are built based on data about the patient’s history and genotype. Secondly, the scientists run a series of simulations testing each drug against each different mutation to see which drug would work best in each case, giving the doctors a better idea of which treatment to use.
The scope of the simulations is enormous, with each one generating up to 15Gb of data. To process this data in the time scales relevant to clinical treatment (generally less than two weeks) the scientists needed to use grid computing services to spread the load of the job across the fastest computers in the UK and the US.
‘The impact of scientific work is only apparent if it can be achieved in time scales suitable of clinical decision making of what the treatment should be,’ Professor Peter Coveney, who led the research, told scientific-computing.com.
However, although it has proved that personalised treatment of HIV may be possible in the future, the research still faces major challenges. ‘It must be tested for robustness and reliability before it becomes a feasible solution,’ says Coveney.
It will also face a number of legal issues to be sure that laws protecting patient confidentiality and data protection are not broken when building the patient-specific models. ‘It would be a disaster if the data become available,’ he says.
The work also raises issues about the availability of grid computing resources for medical applications. Currently, computing jobs are placed in long queues with no order of preference, but in life or death situations this could delay potential treatment until it was too late. In addition, these services may be too costly for most health services.
Coveney predicts that once high-performance computing becomes more mainstream, grid computing services may be seen as a utility service such as electricity and water that is readily available commercially – a solution that could solve these issues.
