Coronavirus: AI Gears Up In Battle Against Covid-19


It feels as if an exceptional effort is needed to help relieve the global pandemic killing so many.

Artificial intelligence may have been overvalued – but when it comes to medicine, it already has a verified track record.

So can machine learning counter this challenge of finding a cure for this dreadful disease?

There is no dearth of companies trying to solve the problem.

Oxford-based Exscientia, the first to place an AI-discovered drug into human trial, is searching through 15,000 drugs kept by the Scripps research institute, in California.

And Healx, a Cambridge company established by Viagra co-inventor Dr David Brown, has repurposed its AI system built to find drugs for rare diseases.

The method is divided into three parts that:

  • search through all the present literature relating to the disease
  • study the DNA and arrangement of the virus
  • consider the appropriateness of various drugs

Drug discovery has conventionally been slow.

“I have been working on this for 45 years and I have given three drugs to market,” said Dr Brown.

But AI is showing much faster.

“It has taken numerous weeks to collect all the data we need and we have even got new data in the last few days, so we are now at a significant mass,” revealed Dr Brown.

“The algorithms ran over Easter and we will have results for the three methods in the coming seven days.”

Healx expects to turn that information into a list of drug candidates by May and at present is in talks with labs to take those predictions into clinical trials.

For those working in the arena of AI drug discovery, there are two choices when it comes to coronavirus:

  • find a completely new drug but wait a couple of years for it to be permitted as safe for use
  • repurpose current drugs

But, Dr Brown said, it was quite unlikely one single drug would be the answer.


And for Healx, that means detailed scrutiny of the eight million possible pairs and 10.5 billion triple-drug combinations stopping from the 4,000 approved drugs on the marketplace.

Microscopic view of influenza virus cells


Prof Ara Darzi, chief of the Institute of Global Health Innovation at Imperial College, said: “AI remains one of our strongest ways to achieve a noticeable solution but there is an essential need for high quality, clean and large data sets.

“Till now, much of this information has been siloed in individual companies such as big pharmaceutical or lost in the intellectual property and old lab space in universities.


“Nowadays more than ever there, is a need to unite these disparate drug discovery data sources to permit AI researchers to apply their novel machine-learning techniques to produce new treatments for Covid-19 as quickly as possible.”


In the US, a collaboration between Northeastern University’s Barabasi Labs, Stanford Network Science Institute, Harvard Medical School, and biotech start-up Schipher Medicine is also on the lookout for drugs that can rapidly be repurposed as Covid-19 treatments.

Amazing findings

Usually, just getting them all to work at the same time would take “a year of paperwork”, said Schipher’s chief executive Alif Saleh.

However, a series of Zoom calls with a “group of people with an unprecedented determination to get things done, not to reveal a lot of time of their hands”, paced things up.

“The last three weeks would usually take half a year. Everyone dropped everything,” he said.

By now, their research has produced surprising results, including:

  • the suggestion the virus may attack brain tissues, which may explain why some people lose their sense of smell or taste)
  • the expectation it may also attack the reproductive system of both men and women

Schipher Medicine integrates AI with something it calls network medicine – a technique that views a disease through the complex interactions among molecular constituents.

“A disease phenotype is seldom due to malfunction of one gene or protein on its own – nature is not that uncomplicated – but the result of a spilling effect in a network of interactions between numerous proteins,” Mr Saleh said.

Using network medicine, AI and a union of the two has led the group to identify 81 potential drugs that could assist.

“AI can do slightly better, not only looking at higher order associations but little bits of independent information that conventional network medicine might miss,” said Prof Albert-Laszlo Barabasi.

Nevertheless, AI alone would not have worked, they required all three approaches.

“Different tools look at different viewpoints but together are very potent” he added.

Some AI companies are already declaring to have isolated drugs that could help.

BenevolentAI has recognized Baricitinib, a drug already accepted for the treatment of rheumatoid arthritis, as a possible treatment to avoid the virus infecting lung cells.

And it has now moved in a controlled trial with the US National Institute of Allergy and Infectious Diseases.

In the meantime, scientists from the US and South Korea using deep learning to explore the potential for commercially available antiviral drugs have recommended atazanavir, used to treat AIDS, could be a good applicant.

Other companies are using AI for other reasons, such as analyzing scans to reduce the burden on radiologists and help forecast which patients are most probably need a ventilator.

Chinese technology giant Alibaba, for instance, announced an algorithm it says can identify cases within 20 seconds, with 96% accuracy.

But some specialists warn AI systems are likely to have been trained on data about progressed infections, making them less effective at diagnosing early signs of the virus.

There needed to be a worldwide effort from policymakers to encourage the big pharmaceutical companies to join hands with smaller drug-data stores, research charities and academics to assemble data resources, Prof Darzi said.

“The time has never been more crucial for drug-discovery data to open up its secrets for AI to assist in the battle against Covid-19,” he said.



Leave a reply