HPC Case Study: Modelling Immuno-Oncology

Using HPC to model the ERAP1 enzyme to better understand autoimmune disease.

No medical technology will ever be as sophisticated as your immune system, the network of lymphatic organs and cells whose daunting task it is to rid your body of viruses, harmful bacteria, and other pathogens. 

But sometimes the immune system goes awry. 

T-cells – the lymphocytes responsible for identifying and destroying infected or cancerous cells – can destroy healthy cells, resulting in a vast range of autoimmune diseases such as type 1 diabetes, multiple sclerosis, and ankylosing spondylitis. Or they can negligently overlook malignant cells, allowing cancer to grow and spread. 

Many of the most consequential medical breakthroughs of the last few years, including the work that led to James P. Allison and Tasuku Honj’s 2018 Nobel Prize in Physiology or Medicine, have belonged to the burgeoning field of immunotherapy, a form of biological therapy that works by activating or suppressing the immune system.

And the University of Southampton’s Centre for Cancer Immunology is at the forefront of this collective endeavour.


Dr Thanos Papakyriakou is a computational chemist from the National Centre For Scientific Research Demokritos in Athens, working in collaboration with Professor Edd James of the University of Southampton’s Faculty of Medicine.

Together, they are seeking to better understand the role of an enzyme called ERAP1 in autoimmune disease. 

ERAP1 is responsible for clipping antigenic peptides before delivering them to the cell surface, whereupon they are inspected by patrolling T cells to establish whether they’re normal. If not, the T cells perform a summary execution, swiftly neutralising the hazard.

Previous research established that ERAP1 is a highly polymorphic enzyme. That means it has many different variations – or allotypes. Different allotypes clip antigenic peptides to different lengths, affecting how the T cells interpret them.

“I could have a different allotype of this enzyme from you,” explains Dr Papakyriakou, “and this gives rise to individual immune responses.”.

It’s imperative to understand how these small changes in these allotypes give rise to different antigenic peptides

Why? Because those variations can result in the development of autoimmune diseases and the spread of cancer. Understanding the cellular pathology of both types of disease may pave the way towards treating or even preventing them.

Dr Thanos Papakyriakou

How HPC Helped

Unfortunately, explains Dr Papakyriakou, “small differences in the amino acid sequences of proteins don’t actually translate to different structures that we can examine experimentally. If I determine the structure of these different allotypes, I will see mostly the same thing”.

In order to study these enzymes – and this goes for all biomolecules – it’s necessary to examine how they move and interact with the molecules and peptides around them.  

And so, says Dr Papakyriakou, “we perform dynamic molecular simulations. That is, computer simulations of the motions of the individual atoms of the enzyme in space and in time”. 

While it would technically be possible to run these simulations on a desktop PC, the timescales would be severely limited. Iridis 5 allowed Dr Papakyriakou and Professor James to obtain much deeper and more accurate data than would have been possible before recent breakthroughs in GPU technology. 

“The more computer power we have, the further we can go in timescales in terms of the motion of biomolecules. We can have a more complete, realistic idea of the system”.

To put that into raw numbers, each of the six available GPUs in the Iridis 5 cluster is 100 times faster than a 16 core desktop PC processor, which means a simulation that would take Iridis 5 a day to run would keep a desktop PC busy for 600 days. 

That’s the time part of the equation, but what about space? 

According to Dr Papakyriakou, Iridis 5 “allows us to simulate larger complexes of biomolecules, larger parts of the cell”. 

Whereas a desktop PC would allow a researcher to examine a small peptide comprising 300 atoms, Iridis 5 would allow them to simulate a large multi-protein complex comprising more than 180,000 atoms. 

Nevertheless, Dr Papakyriakou is quick to point out that even a powerful cluster like Iridis 5 can only simulate the tiniest fraction of a cell – barely enough to justify the word “part” – which comprises millions of molecules.

“We are just simulating a single biomolecule, which is a large molecule that comprises thousands of atoms, but still we are looking at only a very small piece of the reality, which is the cell.

“And the cell is a very small piece of information with regard to the whole organism – the human”.

But this research will inform further research, perhaps most notably in the field of drug discovery. If we set out to create a molecule to inhibit an enzyme, we need to understand how the different allotypes of that enzyme function in different individuals. 

Only an HPC cluster like Iridis 5 can allow us to achieve this understanding.

Accessing HPC

“I have no background in coding at all,” says Dr Papakyriakou. “I’m a chemist”.

You’d think this might pose a problem for a researcher seeking to enlist the aid of an HPC cluster, but Dr Papakyriakou found the process straightforward.

It began with a small change in mindset. “At one point I came to teach myself how to visualise chemical structures, and from chemical structures, biomolecular structures”. 

From there, Dr Papakyriakou was able to take advantage of existing academic software. “I got to be acquainted with computational chemistry – that is, code to calculate the properties of molecules. Some of these properties are also the dynamics of larger systems like DNA and RNA, for which I used academically developed code. I just needed to learn how to apply it”. 

Thanks to the HPC team at the University of Southampton, this was a painless process.

“It was very easy for me to get an account. I had my code installed and running in no time. The software engineers and the other people who take care of Iridis 5 are excellent”.

Dr Thanos Papakyriakou

Iridis 5 is available to every academic at the University of Southampton – even those who are not acquainted with software engineering or the research potential of high performance computing clusters. 

The University employs a dedicated team of HPC Research Software Engineers whose primary purpose is to help scholars to make the best possible use of Southampton’s world class computing facilities and achieve the best possible research outcomes, whatever their field of study. 

For more information, including details on how to get help from the HPC
RSEs free of charge, see: https://rsg.southampton.ac.uk/hpc

Dr Papakyriakou and Professor James’s project is supported by European funded training network Capstone, and their collaboration continues the work they started with Tim Elliott, Kidani Professor of Immuno-Oncology at the University of Oxford.