Artificial Intelligence boost to our understanding of viruses
Published: 05 May 2021
Virus assembly modelling is key to the scientific understanding of viruses and how they behave and multiply. It is at the heart of vaccine science and the quest for effective anti-viral medications.
In a new study, authors Dr Pierre-Philippe Dechant and Professor Yang-Hui He used machine-learning techniques to reveal hidden patterns in virus modelling data from supercomputer simulations.
They show that detailed simulations previously used to model virus assembly have hidden structures that can be identified very efficiently by Artificial Intelligence (AI). This study shows that the results of this intensive computation can be machine-learnt in a matter of minutes to astounding accuracy.
The research Machine-learning a virus assembly fitness landscape has been published in peer-reviewed open access scientific journal PLOS ONE.
It builds on work done by Professor Reidun Twarock and her team at the University of York in the field of Mathematical Virology. Along with Professor Peter Stockley at the Astbury Centre in Leeds, she has pioneered insights into the assembly mechanism of a wide range of viruses.
This approach has revealed that viruses can assemble like self-packing suitcases in a process that is driven by the genetic material of the virus: this acts like a washing line with a series of pegs that attach to the required items. These pegs are called ‘packaging signals’ and help the genome assemble a spherical shell around itself (the virus particle).
The Twarock team simulated this assembly for a simple virus model, which took several weeks on a supercomputer. Data mining this dataset, Dechant and He found that the patterns in the data could be effectively machine-learned. It is hoped that by combining detailed simulations with AI techniques in this way will allow much more detailed virus models to be tackled in the future.
Dr Pierre-Philippe Dechant (Senior Lecturer in Mathematical Sciences, and Data Science Programme Director, York St John University) said: “Twarock and Stockley have shown the power of frequent iterative collaboration between theory and experiment; and our findings show the potential for combining simulations with data science techniques. Enhancing existing methods of research in this way shows the benefits of different specialisms working together to complement each other’s work. This enables us to work faster and gives us an extra tool in the fight to get ahead of viruses - something we’re only too aware of the importance of at the moment.”
Professor Yang-Hui He (Fellow, London Institute, Royal Institution and Professor of Mathematics, City, University of London) said: “The last few years have revolutionised our understanding of fundamental structures in mathematics and mathematical physics by using techniques from modern data science such as machine learning, data mining, and semantic/linguistic analyses. It is really exciting to see the power of such techniques successfully in action in Mathematical Virology as well. The combination of the ongoing AI revolution with the current pandemic makes research along these lines very timely.”
Modelling virus assembly is a core part of understanding how viruses work in order to understand and potentially disrupt them. Once their code is understood, there are also opportunities to steal the virus’ technology for other medical uses such as drug delivery or nanotechnology.
Further information on virus structure can be found in Twarock and Dechant’s recent review paper.