Simulating the future for flu drugs

Two weeks ago, ScienceOmega.com attended the launch of the e-Infrastructure South Centre for Innovation at the Rutherford Appleton Laboratory in Oxfordshire, when the supercomputer Emerald was officially unveiled. One project that has already begun to reap the benefits of Emerald’s high performance computing powers is being carried out by researchers from the University of Bristol.

The team is using the supercomputer to study the loss of effectiveness of antiviral drugs targeted at the highly adaptive H1N1-2009 virus, which caused the first influenza pandemic of the 21^st Century when it emerged in 2009. Neuraminidase is an important enzyme within the virus and as such has been the target for drugs to treat the infection.

I asked Professor Adrian Mulholland from the University of Bristol’s School of Chemistry, who co-leads the study along with Dr Christopher Woods, about the vital role Emerald is playing in advancing their research.

"We are using Emerald to expand on our earlier study, published in the journal Biochemistry," Professor Mulholland explained. "The aim of the project is to understand how mutations of individual residues in the H1N1 "swine flu" virus enzyme neuraminidase lead to resistance to currently available antiviral flu drugs such as oseltamivir (Tamiflu) and zanamivir (Relenza)."

These treatments were previously effective, but their widespread use has resulted in mutations of neuraminidase which are drug-resistant in varying degrees. A study of the mutant swine flu neuraminidase IRHY, for example, suggested that it reduced the effectiveness of oseltamivir by 12,374 times. With the use of a portion of Emerald’s 372 GPUs (graphics processing units), the researchers are making headway in the effort to understand how structural changes in this enzyme can alter the effectiveness of antivirals so dramatically.

"The earlier study examined just two drugs and two mutants, generating about one microsecond of dynamics, and led to very positive preliminary results that suggested the structural mechanisms behind drug resistance.

"Using Emerald, we are able to significantly extend this work, using about one quarter of the available GPUs to study five mutants and four drugs, and are in the process of running over 36 microseconds of dynamics. The simulations should be complete some time in August, with analysis and write-up hopefully completed in early September."

These same long-timescale molecular dynamics simulations of different neuraminidase mutations would take more than 500 years to complete with an ordinary laptop, illustrating the acceleration of discovery to which Emerald is contributing. Preliminary results are already yielding information about the strong hydrogen bonds that are disrupted to transform neuraminidase from the closed form, which is susceptible to the effects of current antivirals, and the open form, which is more difficult for them to bind to.

"We are analysing the simulations using new algorithms and software developed as part of an EPSRC-funded software development project, all of which should give us detailed, statistically significant molecular insights into the origins of drug resistance, together with a 'map' that should help develop new antiviral drugs that are effective against the mutants.

"In effect, we are using Emerald to develop a computational assay to predict whether a particular mutation will reduce the efficacy of a particular drug, and creating computational tools that will help in the development of the next generation of flu drugs."

The hope is that increasing understanding of the molecular interactions between antiviral treatments and neuraminidase will facilitate the development of new drugs which maintain its closed form. Although the approach used by the team has not been applied to other viral enzymes thus far, Professor Mulholland believes it may be useful to consider in future studies.