CASE HISTORIES

High Performance Cluster helps Model Materials for new drug delivery forms.

When you take an aspirin for a headache, I doubt whether you dwell too long on the small white tablet, and consider the design and development that has gone in to its physical attributes.

Research into drug delivery methods is as important as the design of the drug itself. An effective dose must be released at an appropriate rate from a substance that passes regulatory requirements, is stable during storage and transport, and yet is cost effective to produce. Just producing a ‘simple’ white tablet containing the active ingredient together with a binding agent can cause problems during manufacture, where the content never binds but crumbles, or where increased pressure causes tablets to shatter. Computer-based materials modelling is an exciting area of research which is hoped will lead to a quicker time to market benefits for the production of solid dosage forms.

The newly-formed Pfizer Institute is a research collaboration between the University of Cambridge, the Cambridge Crystallographic Data Centre (CCDC) and Pfizer Ltd. The Institute seeks to provide a centre of excellence for the study of pharmaceutical and medical materials by combining the expertise of the University's Department of Materials Science and Metallurgy, the Department of Chemistry and the CCDC with that of Pfizer. Research interests encompass all aspects of the structure, manufacture and behaviour of solid dosage forms, such as tablets, at all relevant scales of operation and use. The research ranges in size scales, modelling the processes of molecular crystallisation through to achieving better powder compaction, tabletting, diffusion and release.

The Materials Modelling Group, led by Dr James Elliott, uses computer simulations to look at tablet structure and manufacture. Cambridge Online worked with the Materials Modelling Group to understand their computational system needs, design an appropriate High Performance Computing cluster within cost constraints, and then implement.

Delivering Cost Effective High Performance.
” The principal modelling techniques we use include molecular dynamics, dissipative particle dynamics and discrete element simulations. These require extensive use of high performance computing facilities for larger systems or more lengthy simulations.” explained Dr James Elliott.

After funding for the Pfizer Institute was confirmed, James had to move fast to put together the team and the systems they would work on. Starting in early summer 2003, everything had to be ready for the new academic year in September 2003. It was clear that the demands of the project meant that they needed their own High Performance Computing (HPC) cluster, rather than using a shared computer facility.

What started the team off on the right systems solution track was seeing an Intel Itanium2 based High Performance Computing evaluation system. The Intel Itanium2 is noted in part for its exceptional floating-point performance. They followed this by talking to Hewlett-Packard and seeing the HP Integrity range of Itanium2-based servers running materials modelling simulations. Although they had found potentially the right platform, the key to a successful system implementation is in the detailed design work. Hewlett-Packard introduced Dr James Elliott to Cambridge Online as the best partner to work with on configuring and supplying the Beowulf class High Performance Computing cluster platform.

Cambridge Online Expertise
Cambridge Online has 25 years experience of supplying IT systems and services. On this project with the Materials Modelling Group, their long-term experience building HPC systems based on both UNIX and Linux meant they could advise on the best configuration for the teams needs. The Cambridge Online's staff are not just technically excellent, but also understand the needs of users in Higher Education and Research. Dr. Richard Bemrose, one of the Cambridge Online Technical Consultants said “My scientific background certainly helps me to empathise with the team’s goals, while my skills with Linux, networks and storage ensure that the system will deliver what they need.”

The final Beowulf class High Performance Computing cluster configuration met the team’s requirements by utilising high performance technology that was both mainstream and available currently. This meant that the components could be sourced in a timely manner to meet the deadlines of the new academic year, and the system would deliver high performance computing within the tight budget. The computational methods used require exceptional floating-point performance, and hence it was decided to use dual Intel Itanium2 processors within the 8 node HP Integrity rx2600 system. These compute nodes are connected through an HP ProCurve gigabit interconnect enabling parallel simulations to aggregate more processing power and effective memory. The turnkey solution was delivered with RedHat Enterprise Linux Advanced Server, Intel C/Fortran compiler suite and message passing interface (MPI) library all pre-installed and configured to the Materials Modelling Group's specifications by Cambridge Online.

The system was built in a custom rack by Cambridge Online at their specialist facility. The Materials Modelling Group office is at the top of a building that only has access via several sets of stairs with tight turns. Thus once the build of the system was completed and checked, the whole thing had to be broken down into pieces that were small enough to be carefully carried up the stairs in protective packaging. Due to careful planning, the rebuild on site proved to be straightforward. As part of the design process, Cambridge Online also included an Uninterruptible Power Supply (UPS). This has already been successfully put to the test, with unexpected power-cuts to the office not causing failure of the system.

Continued progress for the Materials Modelling Group
The initial funding for the Pfizer Institute covers 5 years, and the team has set themselves a range of goals to ensure their success. In the short term, the mixing of hard particles of various shapes and sizes are to be studied on the computer using standard Molecular Dynamics (MD) or Monte Carlo (MC) simulation techniques. The former are particularly attractive, as the effect of dynamic perturbations (such as shear forces) can be examined. The team has adapted the DL_POLY package from Daresbury Laboratory to carry out such simulations in a way that is very similar to the new technique of Dissipative Particle Dynamics (DPD).

Longer term, James and the team plan to extend their methods to look at the ‘flowability’ of powders in different forms, which could help in the development of new methods for producing solid dosage forms. The new HP High Performance Computing cluster will continue to be at the heart of the computation for this new research. “We have been pleased with the system and support from Cambridge Online. By maximizing the build and configuration work at their facility, they minimized any disruption at our offices. Their continued after sales support and expertise has helped us make the most of the High Performance Computing cluster.” said Dr. James Elliott.

BACK