Artificial Intelligence Shows Promise in Fight Against Coronavirus

Algorithms and big data uncover potential coronavirus treatments and vaccines

Key Takeaways

  • AI is helping researchers find ways to battle coronavirus at an unprecedented rate.
  • Scientists are using ultra-fast computers to comb through huge amounts of data to look for treatments.
  • Using AI can generate too much data for scientists to easily comb through.
  • Researchers are optimistic that AI will help find effective treatments.
Doctor in Hazmat Suit watching virtual Hologram Coronavirus
MR.Cole_Photographer / Getty Images 

Researchers are using artificial intelligence (AI) to sift through vast amounts of data to try to uncover potential coronavirus vaccines or treatments and improve diagnoses.

Dozens of companies worldwide hope to accelerate pandemic vaccine and drug development with AI by slashing development processes that normally take decades down to weeks. Their goal is to get a vaccine or therapy for patients as soon as possible. One firm, Biovista, just announced it used AI to identify two already approved drugs that have the potential to combat symptoms of COVID-19. 

“AI can move very quickly and it can analyze and digest very large data sets,” said Justin Stebbing, an oncology professor at Imperial College London in a telephone interview. “It can help us make hypotheses, but it cannot experimentally prove them. We still need laboratory and clinical data.”

Harnessing the processing power of fast computers is only one step towards a vaccine or treatments, researchers say. Finding and testing drugs on humans is a long process that can only be hurried so much.

Faster, but Too Much Data

The pace of research into coronavirus is quickening, thanks in part to AI. For example, Australian biotech firm Vaxine Pty Ltd is moving its COVAX-19 vaccine into final trials in the hopes of approval by the end of the year. Vaxine Research Director Nikolai Petrovsky said in a news release they “aim to collect and analyze the COVAX-19 trial data in real time, rather than waiting until the end of the trial before seeing if the vaccine is working, which is the traditional process.”

The relatively new field of AI-assisted medical discovery faces challenges, however. One problem with using AI is it can produce too much data to easily sift through, said Stebbing. Algorithms are used to trawl huge sets of data to find possible approaches for treatment. But once discovered, those approaches still need to be tested in the laboratory.

Aside from finding new drugs, researchers are also using AI to try to take shortcuts by repurposing drugs already approved. Stebbing co-authored a study that found promising results treating COVID-19 with the rheumatoid arthritis drug baricitinib, along with an anti-HIV drug combination of lopinavir and ritonavir.

“We are going to see another emerging pandemic [and] AI will be important for finding cures.”

“AI helped analyze all the literature and generate hypotheses,” Stebbbing said. “But we still needed to go through all those possible approaches that were identified and rule out the ones that didn’t work or were too toxic.”

Biovista said Cablivi and Atozet could help with blood clotting and inflammation. The company claims its AI platform, Project Prodigy, maps all known drugs against every possible mechanism in coronavirus.

"Typical machine learning AIs aren't designed for something like Covid-19, where what happens next is unknown," said Dr. Aris Persidis, president and co-founder of Biovista in a news release. "Biovista's Project Prodigy is a game-changer AI because it is designed to go where matching and classifying AIs stop. Project Prodigy recombines all known data points to build entirely new scenarios.”

Stebbing said that AI has the potential to revolutionize drug discovery. “This is a very exciting time,” he said. “People talk about bench to bedside research and now this is computer to bedside.” 

Despite Stebbing’s optimism, he cautions that much basic research still needs to be done to understand the virus. Around the world, laboratories are racing to amass data that can help them understand how coronavirus works and how to detect it in patients. 

Computers Search for Better COVID Diagnoses

Using AI to comb through images of diseased lungs could provide another way to diagnose coronavirus. The National Institutes of Health recently launched the Medical Imaging and Data Resource Center (MIDRC), an effort to harness the power of artificial intelligence and medical imaging.

T scan of lungs of COVID-19 patient
T scan of lungs of COVID-19 patient with areas described by radiologists as resembling grains of ground glass. Radiological Society of North America 

One goal of the MIDRC is to help diagnose coronavirus in patients through imaging rather than blood or saliva tests. The team is using algorithms to search through huge databases of pictures of patients' lungs, said Kris Kandarpa, Director of Research Sciences and Strategic Directions at the NIH. In August, Kandarpa’s team began creating a “virtual repository” of images of patients' results from around the world. The FDA is in the process of certifying the AI analysis of this data. 

“What’s been lacking is a large data set,” Kandarpa said in a phone interview. “We are hoping that we will be able to pick up imaging modalities. A person with covid can have a PCR [polymerase chain reaction]negative test and have changes in the lung.” 

The goal of the project is to find a “biomarker” or telltale sign in imaging results that shows a person has COVID. Even if it’s found, the biomarkers won’t be necessarily used to diagnose most cases, Kandarpa said, but could inform epidemiologists about the spread of the disease.

“This is a very exciting time. People talk about bench to bedside research and now this is computer to bedside.” 

Kandarpa said he is cautiously optimistic that “we are expecting that we can get some robust results by the end of the year. I’m not saying it will happen but it's possible.” 

Diagnosing the disease is only one part of the problem. Researchers are also using AI to examine images of the virus itself and try to find its weaknesses.

Like a Key for a Coronavirus Lock

Researchers are also using AI to search for ways to combat the virus at the molecular level. They are hoping to find weaknesses in the virus’s proteins that drugs can latch onto. It’s like searching for the right key to fit in a lock, said Arvind Ramanathan, a computational biologist at Argonne National Laboratory, in a phone interview. 

The challenge is that both the keys and the locks are incredibly small and there are millions of them. That’s where AI comes in—sorting through all the possibilities.  

Argonne National Laboratory’s Theta computer is being used to understand the coronavirus at the molecular level.
Argonne National Laboratory’s Theta computer is being used to understand the coronavirus at the molecular level. Argonne National Laboratory 

“The machine learning code has to learn how to take a small molecule and lock it down,” Ramanathan said. “The only problem is that the lock is flexible. Once you understand the mechanism of the lock you can find a lot of ways that molecules are similar and might work in a drug.”

Once protein targets are found through AI, drugs are tested to see if they will work. But getting drugs to actually cure human patients still remains a challenge. None of the candidate drugs discovered via AI have so far been approved as a treatment. “It’s a very hard problem,” Ramanathan admitted. Despite these challenges, Ramanathan said that he is “quite optimistic” that AI will help discover coronavirus therapies. 

“We have taken a very deliberate approach to how we think about targeting these proteins,” he said. “The fact is that we are bringing many scientific communities together and also I’m optimistic due to the amount of data we have access to today.” 

Using AI will be critical to finding treatments for more than just coronavirus, Ramanathan said. “We are going to see another emerging pandemic,” he said “AI will be important for finding cures.” 

Researchers say that the emerging field of AI-assisted drug discovery is likely to revolutionize medicine. How long that revolution will take, though, is an open question that will hopefully be answered in the coming months.