Cambridge-1 and the future of medicine


The future of artificial intelligence (AI) in medicine is in Cambridge – and it’s blazing fast. Nvidia has unveiled Cambridge-1, the UK’s most powerful supercomputer and among the top 50 in the world, installed during the pandemic at a cost of $ 100million (£ 73million) to the GPU giant.

The goal isn’t just to sweeten Nvidia’s acquisition of UK chipmaker ARM, but to show off its DGX SuperPOD scalable computing setup for AI – and help fuel the future of medicine. “Cambridge-1 will give leading researchers in business and academia the opportunity to do their lifetimes’ work on the UK’s most powerful supercomputer, unlocking clues about diseases and treatments at scale and speed previously impossible in the UK, ”said Jen- Hsun Huang, CEO of Nvidia, at the launch of the new supercomputer.

“The discoveries developed on Cambridge-1 will take shape in the UK, but the impact will be global, leading to groundbreaking research that could benefit millions of people around the world.”

Cambridge-1 is named after the city where genomics and computing made their debut, but it’s also where Nvidia plans to set up its headquarters in the UK. “Cambridge-1, and our purchase of ARM, is our giant investment in the UK, giving us a platform to work with the amazing universities, businesses and over 1,000 AI startups in this community,” said said Huang.

Scalable supercomputing

With eight Linpack performance petaflops, Cambridge-1 is the UK’s fastest supercomputer, narrowly surpassing the previous leader led by the Met Office. How did Nvidia, more famous for its gaming GPUs, achieve this feat? The answer lies two decades ago, when researchers struggling with machine learning began testing Nvidia hardware for their AI projects, sparking an AI boom that continues today. Nvidia has since started developing custom hardware for AI tasks, including the Ampere processor at the heart of Cambridge-1.

The key building block is Nvidia’s artificial intelligence system, the DGX. It includes eight A100 Ampere GPUs as AI accelerators, as well as a 128-core AMD processor, and includes networking and storage as well as software and AI models. The idea is that an organization can easily scale their AI efforts by connecting multiple DGX modules; the SuperPOD that Cambridge-1 is based on stacks 80 together in a single package, for a supercomputer in a box.

What’s intriguing about Cambridge-1 is that it is the first Nvidia supercomputer that has been made available for use outside of the company itself. The chipmaker has already built supercomputers for its own uses, such as developing new GPU technology, but the Cambridge-1 is the first created for outside partners.

However, the Cambridge-1 is not actually in Cambridge. To enable the supercomputer to be powered by renewable energy, it is located at the site of green hosting provider Kao Data in Harlow, Essex.

The future of medicine

AI is increasingly helping medicine in multiple ways, from scanning images to detect symptoms of cancer, to sorting data to identify patterns that can guide research into diseases such as dementia. He also develops models to help create new drugs. However, ethical and safety concerns mean that development must be done with care.

Right now there are “two fundamental problems,” said Sébastien Ourselin, director of the School of Biomedical Engineering and Imaging Sciences at King’s College London. One is access to data, and the other is access to the computing power needed to process it.

On the first issue, industry and the NHS need to work closely together to increase accessibility to relevant data, an area which has been difficult in the UK due to privacy concerns. Ourselin noted that it is critical to ensure that AI is developed using unbiased data sets that represent the demographics of the local population. Otherwise, it won’t work for everyone.

The second problem is growing. The latest AI systems can do amazing things with ever more sophisticated models and massive data sets. But they require a lot more computing resources than older, simpler systems. “The power of your computer limits you,” said Lindsay Edwards, vice president of data science and AI, respiration and immunology, biopharmaceutical R&D at AstraZeneca. “If it takes you three weeks to train a model, your ability to refine that model is really difficult – you have to be able to iterate quickly. “

The power of Cambridge-1 enables scientists to get results much faster than with conventional computers

Hence the eagerness of medical organizations to embark on Cambridge-1. Access to this supercomputer could trigger a leap forward in advanced medical modeling, finally enabling faster breakthroughs in drugs, genomics, and disease understanding.

Early plans to access Cambridge-1 include drug discovery and medical imaging with AstraZeneca, genetics and health predictions with GSK, and genomic sequencing with Oxford Nanopore Technologies. It will also generate synthetic brain images for King’s College London and the Guy’s and St Thomas’ NHS Foundation Trust to fuel research on dementia, stroke, cancer and multiple sclerosis.

“Some of the early projects planned for the supercomputer are using AI to sequence genomics, discover new drugs and unravel the mysteries of dementia by studying MRI scans,” said Huang.

AstraZeneca and drug discovery

AstraZeneca has received a lot of attention lately thanks to its production of the COVID-19 vaccine developed in Oxford, which was being put into arms less than a year after the start of the pandemic – and AstraZeneca is hoping that Cambridge-1 will accelerate future drug development. The pharmaceutical giant is using a new transformer-based neural network architecture to predict drug reactions and optimize the development process. This is a pioneering approach because the network will train efficiently: just as a natural language processing AI will learn the basic rules of language by “reading” as much text as possible, this system aims to learn the rules of chemistry. , explained Edwards.

AstraZeneca is also conducting a trial on digital medical images, hoping to apply a trained algorithm to annotate images of tissue samples, while looking for anything that might signal a drug reaction. “One of the challenges of digital pathology is that the images are huge. The files are incredibly large, ”said Edwards. “What we’ve done in the past is decompose these files and train models on them, but that comes with a whole host of problems. The Cambridge-1 scale computation allows us to process the image of the entire slide, which is really transformational. “

Genetics, GSK and Oxford Nanopore

Elsewhere, GSK and Oxford Nanopore are both working on genetics and genomics projects, hoping to better understand the disease and potential treatments. GSK is using AI to better understand biology using genetic evidence, not only to develop drugs faster, but also to better understand when and how to use treatments. “We know that drugs that are based on genomic evidence are twice as likely to be successful and result in effective drugs for patients,” said Steve Crossan, vice president of AI and learning automatic at GSK.

Meanwhile, Oxford Nanopore is using Cambridge-1 to run a model originally developed for speech recognition, to examine biological markers that are not yet well understood. “There is a tremendous amount of work that is going not only to accelerate the understanding of what these markers are, what you call them and how you make them work in a reproducible fashion, but how they affect health,” said Sinclair Dokos. . The hope, she added, is that access to Cambridge-1 will allow Oxford Nanopore to accelerate the development of its algorithm, leading to more precise genomic analysis.

A Cambridge-1 processor bank

Brain scans for research

King’s College London and the Guy’s and St Thomas’ NHS Foundation Trust are working to better understand brain diseases such as dementia, stroke, multiple sclerosis and cancer, with the aim of accelerating the diagnosis and improve treatment. The model will be trained on tens of thousands of MRI scans to teach her how to simulate brain images with characteristics such as age or disease, giving researchers more material to work on.

“By generating so many images, you can create a situation where you are absolutely certain that you have every opportunity to see what a brain will look like if it is healthy,” Ourselin explained. “But what you could also do is start generating the same amount of data for a specific disease and then start to understand the progression of the disease.”

These are just the first projects: the aim is for Cambridge-1 to bring British AI-based medical research to the forefront. “Cambridge-1 [allows us] to try things that we couldn’t try before, ”said Edwards. It could even help develop the field of personalized medicine, according to Ourselin: “By comparing in-depth information from a specific patient with the entire population, and 20 years or 30 years of data, we will be able to do better accuracy. . medicine, ”he said.

Nadine Hachach-Haram, head of clinical innovation at Guys and St Thomas’, predicts that the discovery of new drugs and treatments will undoubtedly change the pharmaceutical industry. But there’s more: “The even more exciting thing for me is to reimagine healthcare where we bring together cutting-edge life science innovation and computing power, to ultimately allow us to do what we’re trying to do. all to do here: provide the best care to our patients.

Featured Resources

The path to cloud-based innovation

Migrating from SAP to the Cloud Gives Businesses a Competitive Advantage

Free download

Seven main use cases for machine learning

Seven Ways Machine Learning Solves Business Problems

Free download

Drive adoption of digital self-service

From early innovation to mass adoption of digital control

Free download

Three tips for effectively leading hybrid teams

A guide to employee motivation and engagement for business leaders

Free download


About Author

Comments are closed.