By Robert Roe
More robust treatments for antibiotic-resistant bacteria could result from research at the University of Zurich that simulated interactions among bacterial colonies, using an SGI supercomputer there.
The research investigated the cooperation between colonies of bacteria, and could have other applications, including the treatment of Cystic Fibrosis.
Dr Akos Dobay, a post-doctoral researcher for the Institute of Evolutionary Biology and Environmental Studies at the University of Zurich, led the project. Dobay said: ‘It’s already understood that microbes display social behaviour, similar to that of human social groups, by sharing “public goods” at the group level,’ stated Dr Dobay.
Public goods are those available to the bacterial colony as a whole, essentially metabolites that have been secreted and are available within the community. Although public goods generally promote population growth, they are also vulnerable to exploitation by cheating mutants, which no longer contribute, but still benefit from the public goods produced by others.
Understanding these interactions could lead to the treatment of disease by influencing the level of cooperation. Reducing the cooperation in harmful bacteria could limit the population growth, making it easier to treat with anti-bacterial medicine. On the other hand, increasing cooperation could encourage population growth, which could be applied to the production of biofuels.
However, the interaction between so many organisms is complex and requires highly parallel computational simulation.
Dobay said: ‘The study of microbiology can quickly become complicated, even the simplest of microbes are complex and manipulating them isn’t always straightforward. Our research set out to study how bacteria interact and work together in micro colonies under various constraints, and how this contributes to the overall fitness of the colony.
‘We studied four different variables in order to get a significant baseline understanding of how these impact the dynamic of a colony. In microbiology, even four variables can be hugely complicated and time-consuming to study. It would have been near impossible for us to undertake this research using traditional methods, so we developed a specialist program which digitally simulated how microbes interact.
‘With a computational simulation, we can really look the bacteria, how they divide, how they grow, and we can introduce environmental or ecological factors that have an influence on bacterial cooperation.’
The researchers had originally planned to run the simulation on the University of Zurich’s Schrödinger system, a large high-performance computing (HPC) cluster built by Sun Microsystems. However some complications led the team to decide to use the new SGI system called Hydra.
Hydra is an SGI UV 2000 large-memory multiprocessor system that was upgraded in 2013 with 4 terabytes of shared memory and 96 CPU cores (originally it had 48 cores and 512 gigabytes of memory).
Dobay said: ‘The nodes on Schrödinger are connected using quadruple data-rate InfiniBand. When we were planning how to conduct our research project, the InfiniBand was causing a lot of issues for some of my colleagues. Adding to this, Schrödinger was apparently suffering from latency problems, so we didn’t feel confident running our program on something which wasn’t 100 per cent reliable.’
Dobay’s research utilised all of the Hydra system, running the team’s unique simulation program across all 96 cores non-stop for three months.
‘I’ve been really impressed with the reliability of the SGI UV system. This is the first time we have run a research project of this scale across it, and it ran consistently during our study under a heavy workload with no failures and no need for a system shut down, stated Dobay.
Dobay concluded: ‘Our team ran 98 percent of our research on the Hydra SGI system installed at the University, there’s no doubt that without it we wouldn’t have been able to conduct our study, especially in three months.’
The research was published in the Journal of Evolutionary Biology in June this year in a paper, of which Dr Dobay was the first author, entitled ‘Interaction effects of cell diffusion, cell density and public goods properties on the evolution of cooperation in digital microbes.’
The simulations revealed that increased diffusion of cells and public goods selected against cooperation across most of the parameter space, but not when diffusion of public goods was low. In this latter case, co-operators divided more quickly and were consistently favoured regardless of the extent of cell diffusion.
This result can be explained intuitively as lower diffusion of public goods leading to the secreted metabolite travelling less distance from the bacteria that produced it. This results in ‘producers increasingly benefiting from their own molecules.’
However the research found that ‘even with minimal public good and cell diffusion, secreted metabolites remained accessible to cheaters to some extent.’ So ‘complete privatisation’ of secreted metabolites is not possible in this system.
The research also discovered that co-operators divided more quickly and were always favoured when cell diffusion was minimal, ‘regardless of the degree of public good diffusion.’
Dobay said: ‘The key findings of this research was the realisation that one way to escape cheaters was for the co-operators to establish some spatially separated structures, by staying close to each other the co-operators were able to avoid some of the cheating bacteria.’
According to Dobay, a recent extension to the research revealed: ‘If you take a strain that has been previously exposed to cheating and you put it in a colony with naïve cheaters, then you will see that the co-operators will actually do better than if you were placed with cheaters that were more developed.’
Dobay concluded: ‘What was found here is that both, cheaters and co-operators benefit from the presence of the other. You need some enemies in order to become a better soldier so to speak.’