New UK supercomputer boost for life sciences
A new supercomputer built on a public-private partnership in Cambridge is the fastest ever seen in the UK. Powered with 100% renewable energy, it will be dedicated to advanced research in the life sciences.
The behemoth has been built by California-based technology company Nvidia in just 20 weeks, itself a record. Its line of supercomputer processing units, DGX, has been used to power computational science projects, especially deep-learning research, since 2016. The new Cambridge computer connects 80 of the third generation DGX servers, with a total capacity of 400 petaflops. FLOPS stands for FLoating point OPerations per Second and is a measure of computer performance. To put the new machine's performance into perspective, the latest top of the range Playstation 5 console tops out at 10 Teraflops. In other words, it's roughly the equivalent of 10,000 Playstations worth of processing power, although in truth, thanks to its architecture and programming, it's much more powerful than that. Nvidia has invested around £73 million in the project, which they've named Cambridge-1. The company expect to turn a profit of some £600 million over the next 10 years.
The second part of this project involves the development of an Artificial Intelligence centre of excellence in Cambridge. This will help with research fellowships and partnerships among scientists from across the country who need scalable AI-specific computing capability. There will also be a start-up accelerator, comprising privately-funded student programmes including entrepreneurship mentorships to pitch business ideas to potential investors. At the moment, Nvidia is in the bidding process of acquiring the most important chip maker in the UK, ARM, that could power the next generation of supercomputers that are being designed in the Cambridge cluster of high-tech businesses known as ‘Silicon Fen’.
The CEO of Nvidia, Taiwanese-born technology entrepreneur Jensen Huang, released a video for the launch event. “Cambridge-1 will empower world-leading researchers in business and academia with the ability to perform their life’s work on the UK’s most powerful supercomputer, unlocking clues to disease and treatments at a scale and speed previously impossible in the UK," predicts Huang.
But, is it all about speed? And what will be the real uses of this computer? Currently, five parallel scientific projects are taking up most of the capacity, using not only the speed but also the system's ability to run complex processes. This is made possible by using a scalable server architecture that dynamically engages and disengages processors to make more computer power available as required.
Pharma giant AstraZeneca, whose headquarters are in Cambridge, is already testing the potential of Cambridge-1 for drug discovery. They're running a new model, named ‘MegaMoIBART’, that automates reaction predictions and molecular optimisation. The mathematical model uses self-supervised processes capable of predicting outcomes based on previous analyses of large datasets. The model, using the supercomputer capacity, will suggest new therapeutic molecules based on long training sessions reviewing scenarios of previously analysed datasets.
GlaxoSmithKline’s (GSK) main use is on analysis of genetic data for predictive research on drug discovery. In this case, speed is key and projects will benefit from focusing the efforts on those studies that are more likely to succeed.
King’s College London and Guy’s and St Thomas’ NHS Foundation Trust are training AI models by exposing them to datasets of MRI brain scans so they can generate synthetic brain images. These images could be used to increase the understanding of brain diseases and facilitate faster diagnosis and treatment. It will open new partnerships and collaboration with the NHS and biomedical databases such as the UK Biobank. Professor Ian Abbs, CEO of Guy’s and St Thomas’ NHS Foundation Trust explained: “The power of artificial intelligence in healthcare will help to speed up diagnosis for patients, improve services such as breast cancer screening, and support the way that we risk assess and prioritise patients according to clinical need”.
Oxford Nanopore Technologies provides DNA and RNA sequencing products that are used in fields ranging from healthcare to environmental monitoring. They will be able to reduce the duration of algorithm processing from days to hours.
In summary, Cambridge-1 has been built in record time, will be focused on AI-driven scientific projects, and is the fastest supercomputer currently in the UK. It could be the beginning of a race to build more complex ones with Cambridge as the central hub for development and computer power distribution to other research centres in Europe.