SINGAPORE – The Republic is building its third national supercomputer to advance research in Covid-19 and other areas in healthcare.
The new supercomputer is so fast it is expected to train artificial intelligence (AI) to predict a patient’s future disease condition – such as Covid-19 and kidney disease – within hours, instead ofdays as with standard computers.
The past data of patients with specific diseases can be fed into the supercomputer to train the AI.
The training can be done more quickly because the supercomputer can carry out petascale computing. This means it can perform more than a quadrillion – or over a thousand trillion – calculations in a second.
Other coronavirus research that can benefit from the supercomputer includes figuring out drug combinations for treating Covid-19 infections as well as discovering new drugs.
The National Supercomputing Centre Singapore (NSCC) and National University Health System (NUHS) finalised an agreement on Friday (Dec 3) to build this third national petascale supercomputer, which is expected to be ready by the middle of next year.
“(The supercomputer) enhances the speed at which we can develop new therapies and care paradigms, like being able to predict patients’ conditions,” said Associate Professor Ngiam Kee Yuan, NUHS’ group chief technology officer .
Referring to AI trained by the machine, he said: “At a very basic level, AI will be able to calculate the risk of (hospital) readmissions. This is not something a human doctor can easily give a number to. Taking that further, if a patient has a set of conditions, the AI is able to predict what’s going to happen in the future to these conditions.”
Unlike the first two national supercomputers, the third one, which has no name yet, is dedicated to healthcare and medical research.
The first one, called Aspire 1, was set up in 2016, while the second, Aspire 2a, is expected to be operational by early next year. Their use covers a broad range of research areas such as climate change, weather monitoring, urban planning, healthcare and materials research.
The third will be sited in NUHS, but as a national supercomputer, it will be open for Singapore’s healthcare sector to use.
One advantage the new supercomputer will have over the others is that being located at NUHS, clinicians using it will not have to send the data to be analysed to another supercomputer located elsewhere.
This means the extra step of processing patients’ data is not needed and it can be analysed more quickly.
But NUHS assured the public that patients’ information analysed at the new supercomputer will be anonymised to safeguard people’s privacy. NSCC added that the data will be protected within NUHS’ secured environment.
The cost of building the third supercomputer was not immediately available, but Aspire 2a costs $40 million to build.
The funds for the new supercomputer will come from the $19 billion that the Government set aside to develop new technologies under the Research, Innovation and Enterprise 2020 plan,
Prof Ngiam said that the new supercomputer is also expected to help identify genetic variations when studying genes and for the development of AI imaging tools such as to identify spinal cord issues.
The supercomputer can be used to train other systems too. One is the AI that helps a machine identify objects in common places, such as along a hospital corridor. Another is a chatbot that can have conversations with patients in a more natural way.