Will AI be the most powerful tool against coronavirus – COVID-19?

AI in the fight against coronavirus

Date: 12th February 2020

Scientists and biotech companies are coming together in the race to fight the novel coronavirus (COVID-19) – a zoonotic coronavirus that has jumped from infected animals to humans.  Whilst, there is no substitute for epidemiologists, scientists and public health officials, AI is poised to play a crucial role in multiple areas of controlling and treating the disease. Here we summarise some of the key areas where AI has played a crucial role so far and where else it might help.

Yesterday the virus responsible for this devastating outbreak finally got a name, the World Health Organization (WHO) have proposed COVID-19, standing for coronavirus disease 2019. With, 40 554 confirmed cases to date, over 900 deaths, and 24 countries outside China having confirmed cases the key to minimising the devastating effects of the virus is speed.  This is where AI excels, its ability to compile rapidly evolving information, analyse large data sets, combined with learning capabilities of the algorithms makes AI a hugely valuable tool.

AI-driven algorithm discovered the coronavirus outbreak

The WHO first reported the potential outbreak of pneumonia of unknown cause in China on the 5th January, 2020, however it was a Canadian company, BlueDot that first alerted the authorities (and its customers) of the potential outbreak.

BlueDot, uses AI solutions to track, contextualise, and anticipate infectious disease risks.  Using natural-language processing and machine-learning techniques the algorithms sift through global news and official reports, including animal and plant disease networks.  They anticipated the potential risk of the coronavirus making their first alert on the 31st December 2019.

Furthermore, using historic airline data, the founder of BlueDot, Kamran Khan, and his team evaluated the potential for international dissemination of this disease via air travel, which later proved to successfully identify Bangkok, Seoul, Taipei and Tokyo as areas of risk.

AI-driven epidemiology

The ability of AI to predict the initial spread of coronavirus via air travel is likely to evolve yet further.  In the upcoming months it has the potential to predict how the virus is going to spread based on environmental conditions, access to healthcare, and the way it is transmitted. Perhaps identifying hot spots of infections thereby allowing use to predict where to concentrate our efforts in healthcare resources.

AI could also identify and find commonalities within these hot spots, and in the same vein discover commonalties in cases presenting with similar severity to provide insights into clinically relevant information. The insights from these events may help us answer many of the unknown questions about the nature of the virus.


AI-driven treatment

Understanding the spread of the disease, how it is transmitted and incubation periods are crucial to limiting exposure to the virus.  AI will almost certainly play an accelerant role in finding effective cures or treatments.

This week has seen two papers published in Nature, in which scientists have examined the complete genomes of coronavirus samples collected from patients at the early stage of the outbreak, this will give us vital clues on treatment strategies going forward.

One team, from Wuhan Institute of Virolog, led by Zheng-Li Shi, showed that samples were nearly identical to each other, and shared 79.5% sequence identity to SARS-CoV.  Furthermore, COVID-19 uses the same cell entry receptor as SARS. Analysis at the whole-genome level showed coronavirus was 96% identical to a bat coronavirus, suggesting the virus originated from bats.

The second team, led by Yong-Zhen Zhang from Fudan University, China, also revealed that the virus was most closely related (89.1% nucleotide similarity) to a group of SARS-like coronaviruses, and again bats seem to be the origin,

The similarity to SARS may be of benefit as treatments that have been developed for SARS may prove effective for this current coronavirus outbreak.  However, this is far from certain, and current treatment such as antivirals and vaccines have not been extensively tested.  Furthermore, there is no guarantee they will prove effective. As such there are huge efforts into finding a ‘cure’ or treatment for COVID-19 itself, once again AI is already being used to accelerate the search.

AI-driven vaccines

Many companies have announced their intention of developing vaccines against this novel coronavirus.  We have recently seen the power of AI-driven cancer vaccine as the world’s first enters clinical trials.  Here, it was estimated that the use of AI sped the process up over 4 times, however, whilst impressive, it still took 12 months.  Therefore, it is likely that vaccines are not a short-term solution.  They could still remain an important tool in controlling this epidemic later down the line.


coronavirus treatment

AI-repurposing of drugs

Another AI-approach which may prove to be the most fruitful is to repurpose already existing drugs.  Last week saw four publications based around this theory (three articles were preprint and not certified by peer review).

The first was published on BioRxiv, led by Keunsoo Kang , from Dankook University, South Korea.  Through a drug-target interaction deep learning model they predicted commercially available antiviral drugs that may act on the novel coronavirus.  The result showed that atazanavir, an antiretroviral medication used to treat and prevent the human immunodeficiency virus (HIV), was the best candidate chemical compound.

The next was published in the peer-reviewed The Lancet.  Here Justin Stebbing and his team used BenevolentAI’s knowledge graph -a large repository of structured medical information connected by machine learning – to search for drugs that might block the viral infection.  The team identified baricitinib, which was predicted to reduce the ability of the virus to infect lung cells.

The third was published on MedRxiv led by Feixiong Cheng, from Cleveland Clinic, US.  The team used an integrated network-based systems pharmacology methodology for the rapid identification of repurposable drugs and drug combinations for the potential treatment of COVID-19. The researchers showcased three potential drug combinations (including sirolimus plus dactinomycin, mercaptopurine plus melatonin, and toremifene plus emodin).

The last was reported in Preprints.  Here, Yanjie Wei and colleagues from Shenzhen Institutes of Advanced Technology, China, have used deep learning based drug screens for COVID-19.  Coronavirus main protease is considered to be a major therapeutic target, therefore the scientists focused on drug screening based on the modelled COVID-19 main protease structure.   The deep learning based method DFCNN, could identify and rank the protein-ligand interactions with relatively high accuracy, and was capable of performing virtual screening against 4 chemical compound databases (including against on-market drugs).

The four approaches were designed to minimise the translation gap between preclinical testing and bringing the drug to the clinic.  In epidemic type situations, where time is of the essence this circumvents significant problems in the rapid development of new efficient treatments.  Whilst, these drugs are far from being tested, and in many cases there is no real-world evidence supporting the prediction, it gives a ‘best’ in chance list of drug from which to focus efforts for further validation.

AI-driven novel drugs

Whilst repurposing drugs may save time, there are efforts also being made to use AI in the design of novel drugs.

Insilico Medicine has decided to utilise a part of its generative chemistry pipeline to design novel drug-like inhibitors of COVID-19 using generative deep learning approaches. They have used three of their previously validated generative chemistry approaches: crystal-derived pocked-based generator, homology modelling-based generation, and ligand-based generation.  In their paper published on ChemRxiv, Yan Ivanenkov and the team used 28 machine learning models in the generative phase of the drug discovery process, taking only four days.  They described 6 molecules in the paper which represent drugs from the first batch of generation.  Insilico’s aim – to synthesise and test up to 100 molecules however, recognising the urgency of the situation they are seeking collaborations to synthesise, test, and, if needed, optimise the published molecules.

AI robot medical

AI-bots and screening

AI-backed technology is also playing an important role outside research in efforts to limit the spread of the disease.  From AI-robots in hospitals in the province of Guangdong, which deliver food, medical supplies and can clean wards and themselves, to AI voice assistants giving advice on home quarantine.

Meanwhile, in Singapore, the iThermo, an AI-powered system by Integrated Health Information System (IHiS) and healthcare technology company Kronikare is being piloted as a rapid temperature screen, able to screen people up to 5 times faster than manually possible.

In similar efforts subway passengers in Bejing, China, are being screened en masse for symptoms of coronavirus by AI-temperature scanners developed by two Chinese AI giants, Megvii and Baidu.


With the number of coronavirus cases steadily rising, the world’s scientists have mobilised and are collaborating in efforts to find rapid treatments.  Indeed, as this goes to press the WHO is convening a global research and innovation forum in Geneva to mobilise international action in response to the virus. Their aim is to harness the power of science which will be critical in fighting this outbreak.  They will fast-track the development and evaluation of effective diagnostic tests, vaccines and medicines.  In this respect AI is perfectly poised to accelerate these areas.

By embracing AI it is hoped the effects will be far reaching, from predicting the spread of infections, using data to understand the nature of the virus, to aiding complex decision making.  AI-driven vaccines and drug discovery could vastly increase the rate of effective treatment development and implementation.   It is essential the flow of information and data is open and shared, and it certainly seems that is the ethos of most.  Together, it is hoped we will combat the disease.