SireAdvanced Molecular Simulation Framework

There's been a small delay with publication - we only received the proofs this weekend, so unfortunately the paper didn't make it into a November JCP. Hopefully it should make it into a December JCP so that it is published in 2007. Speaking of papers, I am currently finishing off a review of multiscale modelling of biomolecular systems. There's been a lot of really clever ideas and progress over the last 2-3 years in this field, and it is now possible to perform some really cool simulations that capture a very wide range of length and time scales.

I was incredibly busy last month getting Sire ready for the Supercomputing 2007 conference in Reno. I was very lucky to be given almost complete access to a large rack of ClearSpeed CATS nodes. I had 10 nodes all to myself, each of which contained 12 ClearSpeed accelerator cards, which together were capable of 1 TFLOP each (so I had about 10 TFLOPs to play with!). I was (and indeed still am - ClearSpeed have left the machine on for us to continue running on) running binding free energy simulations on neuraminidase inhibitors using a fully QM model of the inhibitor (Molpro was performing the QM calculation over all of the ClearSpeed cards). This is really cutting edge stuff, and, computer-geek as I am, I found it so cool to be remotely logging in to this unique and first-of-its-type machine and remote controlling it from the other side of the planet. Over the conference we ran for about 1.5 days, and performed something like 10^16 double precision floating point calculations (some of which were on battery, as the entire rack ran off a UPS while there was a power cut - which itself is cool - a battery powered 10 TFLOP supercomputer!).

As well as the first protein simulation, Sire also achieved another milestone last month. It now has its first official user! A project student in the group is now using Sire to run the simulations for a project that is looking at auto-parameterising MM forcefields. It is quite interesting, as the project student has not previously used any other molecular modelling programs, so Sire is their first introduction. One of my design goals is that Sire should be easy to manipulate, and relatively straight-forward to understand. Given the student's excellent progress, I think I may have got the level just about right (and I am now more confident that I can eventually achieve my ambition of building a research group that will all use Sire for their research).

I am now really, really busy. I've gone past my initial deadline for the multiscale review paper, so am now rushing to finish it so that I don't go too far beyond the deadline. I also have quite a bit of coding to do on Sire this month to get it fully ready for some production-quality protein simulations. I have several major pressures to get Sire ready - the success of the ClearSpeed demo in November means that we really want to get the production-ready simulations run as soon as possible (ideally January-March). There is a PhD student who is scheduled to use Sire from January for QM/MM scoring calculations. And, in other exciting news, I have been successful in an application for European supercomputing time (with collaborators at Bristol and at Wroclaw, Poland), where a fair chunk of the time will be used to run computational enzymology simulations (QM/MM free energy) using Sire. I am not complaining, but am definitely looking forward to the Christmas holidays... :-)