Getting to grips with molecules
The Software Sustainability Institute helped a turn a one-man, small-scale software project into a successful multi-developer programme that is transforming research into molecular binding.
The Adaptive Multi-Resolution Massively-Multicore Hybrid Dynamics (AMRMMHD) project, funded by EPSRC, was part of a group of projects surrounding Sire, a molecular simulation framework developed by Christopher Woods from the University of Bristol.
The project improved the efficiency of dynamics sampling algorithms used in the software, and enabled them to run on a wider range of computer resources including the BlueCrystal machine at Bristol and the Emerald GPGPU machine at Rutherford Appleton Laboratory.
"I work on predicting whether molecules will stick together - whether a drug will stick to a protein, for example - and just trying to find better ways to see how things move and bind," Woods says.
AMRMMHD had two aims. The first was to use dynamic algorithms to speed up the simulations, which previously took a long time to run. As Woods says, "we were aiming to calculate the binding affinity of a drug to a protein in a day rather than a week, for example". Secondly, the project team split the code into independent 'chunks' that could be shared across a variety of compute resources, including large clusters and on graphics processors. This increases performance enormously, transforming how the software could be used.
The pharmaceutical industry is constantly on the look out for new and better ways of modelling proteins, for faster drug discovery. The codes developed by AMRMMHD have already been used to study the drug resistance mechanisms of influenza strains, including this year’s outbreaks. Without the new software and algorithms developed by Woods and his team, this simply could not have been done quickly enough.
At the beginning, however, Woods found the project challenging. "I had been working on Sire for seven or eight years on my own, when funding for the AMRMMHD project gave me the opportunity to work with another post-doc and really take the software forward," he says. Yet it was a tough transition from working on his own to collaborating and communicating with someone new, and Woods says he would have really struggled without help.
The Institute "helped us to define just how we would work together, how to break the planned software development into parts and ensure that it remained compatible. They taught me how to become a software manager rather than just a developer," he says. Woods also learned how to communicate his ideas on the structure and the design of the project to fellow developer Ben Long, who was recruited as a post-doc after completing a PhD in Computer Science at the University of Bristol.
Neil Chue Hong, director of the Software Sustainability Institute, met with Woods and Long early in the process and worked through the whole project to identify what could be done and by whom.
"We sat with a whiteboard and identified the interface between our areas. That was so good - it really let me let go," Woods says. "The code by this stage was huge, and I knew every line. To let someone else come in, without constantly reviewing their code, was really hard - but now I could trust Ben, and eventually other developers from the University of Edinburgh who joined in, because I knew we had a language and an understanding in common."
The AMRMMHD project was funded in two stages. The first stage created a prototype, and the second shared it more widely as a cloud-based service. Unfortunately, funding was not continued for Stage 2 - but here again, the Software Sustainability Institute was able to offer its help.
"We've closed it down into a self-contained, well-documented state so that it can be picked up again in future," Woods says. "Neil showed us how we could store all of the knowledge that Ben has gathered, so that it's there for anyone who works on it in future.
"The management and community side to what the Institute does is what makes it stand out for me. The ability to offer that help makes them different - I can't speak highly enough of them," he concludes.