The role of AI in defeating COVID-19
Given the ever-expanding scale and rapid speed at which coronavirus is spreading, the role of AI in defeating COVID-19 has become a focal point for the life sciences sector. Artificial intelligence (AI) tools used in healthcare are set to uncover possibilities for patients, doctors, and researchers, enabling them to make more informed decisions and achieve better outcomes both in their fight against COVID-19 and in the future of healthcare.
From bringing mental health treatment to our fingertips, to predicting when patients recover from a coma, artificial intelligence technology has recently been one of the biggest life-changing trends in healthcare. Until recently, every other healthcare headline on innovations was related to incorporating AI and machine learning – an algorithmic system that improves and adapts through iterative experience without being explicitly programmed, mimicking the human capacity to learn.
This technology has proven to be useful in discovering ways of improving healthcare solutions and assisting medical professionals in accelerating various drug development processes. That said, there is still a question mark looming over AI’s utility in the face of a global pandemic and its unpredictably disruptive impact.
When considering this approach in targeting a serious infectious disease outbreak, AI systems have several limitations that drive scepticism amongst medical professionals. One key challenge is the over reliance on historical training data. AI systems typically need access to large samples of data to correctly identify relevant patterns. The absence of this may lead to an unintentional bias in its future projections stemming from inherently skewed or incomplete data. Additionally, AI processes assume that conditions at present are the same as the conditions characterised in the training data – this sets a certain limitation on AI’s predictive power given the numerous nuances of this new pandemic.
Having said that, AI has the potential to sift through training data and trace patterns and insight, previously overlooked by humans. This can result in the identification of novel solutions for targeting disease using historically available information.
Monitoring and detecting tool
It has been revealed that a Canadian start-up, BlueDot, was the first organisation to report the spread of the coronavirus, beating the World Health Organization (WHO) and the US Center for Disease Control and Prevention (CDC) by nine whole days. BlueDot leverages machine learning and natural language processing to track and recognise foreign-language news reports, animal and plant populations, disease networks, and official epidemiologist announcements to report the emergence of new disease outbreaks, such as COVID-19 in Wuhan, China.
Testing, testing, testing
Though several trials for candidate vaccines and potential therapies are underway, there is currently no cure. In such an environment, diagnostic testing is a key step towards easing the extreme isolation protocols introduced by most countries.
On May 11th, it was reported that researchers at King’s College London (KCL), Massachusetts General Hospital and health science company ZOE had developed an AI diagnostic tool that could predict whether a person is likely to have COVID-19, based on their symptoms. These findings are published in Nature Medicine.
The AI model uses data from the COVID Symptom Study app by comparing people’s symptoms and the results of currently available tests. Two clinical trials in the UK and the US are due to start shortly.
As reported by KCL, the researchers analysed data gathered from c.2.5 million people in the UK and US who had been regularly logging their health status in the app, around a third of whom had logged symptoms associated with COVID-19. Of these, 18,374 reported having had a test for coronavirus, with 7,178 people testing positive.
The researchers created a mathematical model that predicted with c.80% accuracy whether an individual is likely to have COVID-19 based on their age, sex and a combination of four key symptoms: loss of smell or taste, severe or persistent cough, fatigue and skipping meals.
Widespread adoption of the app could help to identify individuals who are likely to be infectious as soon as the earliest symptoms start to appear, focusing tracking and testing efforts where they are most needed – according to the study.
Similarly, the researchers at the University of Cambridge (UK) have turned their attention to AI algorithms to study COVID-19 patient information from Public Health England and predict the risk of patients with developing more severe disease symptoms.
The model uses existing data on hospital admissions, ICU admissions, use of ventilators, patient outcomes to answer questions such as: Which patients are most likely to need ventilators within a week? How many free ICU beds is this hospital likely to have in a week? Which of these two patients will get the most benefit from going on a ventilator today?
Clinical trials and drug discovery
On March 31st, a UK-based leading AI driven drug discovery company, Exscientia, announced a joint initiative with UK’s Diamond Light Source and Calibr, a division of US-based Scripps Research to identify COVID-19 drugs.
Through this alliance, Exscientia has gained access to Calibr’s world’s leading collection of 15,000 clinically ready molecules, including launched drugs and additional compounds currently in clinical and pre-clinical trials. Exscientia will apply its advanced biosensor platforms to rapidly screen the complete collection against key viral drug targets of SARS-CoV-2, the virus responsible for COVID-19.
In February, another UK-based company, BenevolentAI, launched an investigation using its drug discovery platform, The Benevolent Platform®, to identify approved drugs which could potentially stop the progression of COVID-19 and “inhibit the cytokine storm and reduce the inflammatory damage associated with this disease”. According to the company press release as of April 10, 2020, BenevolentAI reported its research findings in The Lancet and again twice in Lancet Infectious Diseases proposing baricitinib as a potential treatment for COVID-19 symptoms and now in clinical trial for this indication.
Baricitinib is an approved drug developed by Eli Lilly and Incyte for the treatment of rheumatoid arthritis. Elli Lily reports that given the inflammatory cascade seen in COVID-19, baricitinib’s anti-inflammatory activity has been hypothesised to have a potential beneficial effect in COVID-19 patients admitted to hospital prior to the development of critical lung damage. A randomised Phase 2 trial with abti-ang2 in COVID-19 set up by Eli Lilly with the US National Institute for Allergies and Infectious Diseases (NIAID) began in early April.
On May 11, a clinical trial sponsored by the NIAID in the US was launched to assess remdesivir (read Result’s report here) plus baricitinib for the treatment of COVID-19. The new NIAID trial, named ACTT2, follows the ACTT study that assessed remdesivir alone in COVID-19 patients. NIAID director Anthony Fauci said: “We now have solid data showing that remdesivir diminishes to a modest degree the time to recovery for people hospitalised with COVID-19…ACTT 2 will examine if adding an anti-inflammatory agent to the remdesivir regimen can provide additional benefit for patients, including improving mortality outcomes.” The main objective of this combination trial is to compare time to recovery between remdesivir alone and in combination with baricitinib.
Industry giants unite in their battle against COVID-19
In addition to AI focused start-ups, government organisations, and universities, the majority of the well-known giants of the 21st century are racing to battle COVID-19. On March 22nd, IBM, Amazon, Google and Microsoft announced a partnership with the US White House to provide computer resources for COVID-19 research.
Google’s DeepMind hopes to contribute to the scientific effort using the latest version of their AlphaFold system by releasing structure predictions of several under-studied proteins associated with SARS-CoV-2. Predicting a 3D molecular structure of a virus and its functions can be crucial information for scientists working to limit the spread of the disease. In theory, without having access to advanced AI technologies, testing all the structural possibilities of a molecule would take longer than the current age of the universe (see Levinthal’s paradox).
Similarly, Microsoft’s John Kahan, Microsoft Global Head, AI for Health Program, notes that Microsoft wants to “make sure researchers working to combat COVID-19 have access to the tools they need” by expanding access to the Azure cloud and High Performance Computing capabilities.
Conclusion
There are several highly efficient start-up companies that are challenging conventional approaches in targeting infectious and other diseases through their focus on AI technologies. Although, some experts warn that AI systems are less effective at detecting early signs of the virus, there is ample evidence that a combination of human creativity, AI technology and healthy competition might play a key role in speeding up a global pandemic recovery. That said, please do stay at home for now, protect the healthcare system and save lives!
AI companies involved in COVID-19 research
The list below is not exhaustive; information is based on company websites and publicly available information. If you’d prefer to download this list you can do so here.
| Company | Area | Location |
|---|---|---|
Benevolent AI Benevolent AI’s knowledge pipeline pulls data from various structured and unstructured biomedical data sources and curates and standardises this knowledge via a data fabric. This is fed into their proprietary knowledge graph which extracts and contextualises the relevant information and is made up of a vast number of machine curated relationships between diseases, genes, and drugs | Repurposing existing drugs | UK |
Innoplexus Innoplexus provides advanced AI and blockchain solutions that support all stages of drug development from pipeline to market. Structured and unstructured life science data is identified by scanning up to 95% of the world-wide-web and merges it with enterprise and third-party data in an ongoing, real-time process. This continually updated data repository is the foundation of their custom, off-the-shelf solutions. | Repurposing existing drugs | Germany |
Gero Gero aims to discover new drugs targeting complex diseases using their next-generation artificial intelligence platform. Gero has managed to overcome limitations of the previous-generation AI by offering not just a correlation analysis of biological big data, but causative models built with the use of physics of complex dynamic systems in addition to advanced machine learning. | Repurposing existing drugs | Singapore |
Cyclica Cyclica’s proprietary structure-based and AI-augmented drug discovery platform includes Ligand Design for multi-targeted and multi-objective drug design, and Ligand Express for off-target profiling, systems biology linkages, and structural pharmacogenomic insights. Taken together, advanced lead-like molecules that minimize unwanted off-target effects are designed, while providing a holistic understanding of a molecule’s activity through integrated systems biology and structural pharmacogenomics. | Repurposing existing drugs | Canada |
Healx Healx’s AI platform “Healnet” uses natural language processing (NLP) to extract disease knowledge from published biomedical databases & proprietary, curated data. Healnet’s data is integrated into the largest, rare disease-focused Knowledge Graph revealing prioritised hidden and novel connections between drugs and diseases. Their approach is hypothesis-free: Healnet leverages the breadth of its disease and drug-specific data, to deliver unparalleled insights. | Repurposing existing drugs | UK |
Vant Vant AI algorithmically generates and optimizes novel drug designs, reverse engineered from target and disease biology. Its AI-powered MOA allows understanding down to the atomic level enabling Rapid and accurate modality agnostic target prediction. Their systematic mapping of drug-target combinations to pathology phenotypes and molecular pathways shows potential for drug repositioning and adverse event prediction. | Repurposing existing drugs | US |
Deargen Dr. UG is an AI platform for developing new drugs that is based on DEARGEN’s deep learning technology. It specializes in genome data analysis, biomarker prediction, molecule selection and optimization. DEARGEN discovers new disease targets with Dr.UG and further realizes precision medicine by designing small molecules for new drugs. | Repurposing existing drugs | Korea |
Exscientia Exscientia’s proprietary AI drug discovery platform “Centaur Chemist™” has reduced hundreds of compounds analysed per drug, dramatically reducing the time and cost of medical discovery. Exscientia is the first AI company to successfully design small molecule compounds that have reached clinical trials. Exscientia’s platform is able to flexibly design and evolve single molecules that hit multiple targets, effective against highly networked diseases. They hope to transform drug discovery into a formalized step whilst allowing the system to learn from human experts. | Designing new drugs | UK |
Insilico Insilico Medicine has developed a comprehensive drug discovery engine, which utilizes millions of samples and multiple data types to discover signatures of disease and identify the most promising targets for billions of molecules that already exist or can be generated de novo with the desired set of parameters. | Designing new drugs | Hong Kong |
SRI SRI’s end-to-end SynFini platform automates the design, reaction screening and optimization, and production of target molecules. It was developed to bring new drugs to clinic faster and at lower cost by accelerating chemical discovery and development. The SynFini closed-loop system comprises of three components that work seamlessly together: a software platform (SynRoute™), a reaction screening platform (SynJet™), and a multi-step flow chemistry automation and development platform (AutoSyn™). | Designing new drugs | USA |
Iktos Iktos’ AI technology, based on deep generative models, bring speed and efficiency to the drug discovery by auto-designing virtual novel molecules that have desirable characteristics of novel drug candidates. This allows for rapid and iterative molecule identification and validation filtering for multiple bioactive attributes and clinical testing criterias. Iktos offers its technology both as professional services and as the SaaS software platform, Makya™. Iktos is also developing Spaya™, a synthesis planning software based upon Iktos’s proprietary AI technology for retrosynthesis. | Designing new drugs | France |
BlueDot BlueDot is proprietary SAAS designed to locate, track, and predict infectious disease spread. BlueDot’s engine gathers global data on 150+ diseases searching every 15 minutes, 24/7. Their predictive ability comes from official data from organizations like the WHO as well as unstructured data including: the annual movements of 4+billion travellers on commercial flights; animal populations; satellite climate data; local journalist and healthcare info and 100,000+ online articles a day. Data is manually taxonomically classified, and the system is trained with machine learning and neural linguistic programming. | Disease surveillance | Canada |
DarwinAI DarwinAI has developed and open-sourced COVID-Net, a convolutional neural network for detecting COVID-19 through chest radiography. Darwin AI is also making the underlying dataset available. The dataset, called COVIDx, comprises nearly 16,756 chest radiography images from nearly 14,000 patients. The team is also accepting submissions to add to its dataset via e-mail (addresses are listed on the GitHub page). Experiment results showed that COVID-Net can detect COVID-19 infection with a positive predictive value (PPV) of 88.9 percent. | CT scan identification | Canada |
SOURCES:
“AI CAN HELP WITH THE COVID-19 CRISIS – BUT THE RIGHT HUMAN INPUT IS KEY”, MARCH 30, 2020, BY MATISSA HOLLISTER, WORLD ECONOMIC FORUM; HTTPS://WWW.WEFORUM.ORG/AGENDA/2020/03/COVID-19-CRISIS-ARTIFICIAL-INTELLIGENCE-CREATIVITY/
“BARICITINIB AS POTENTIAL TREATMENT FOR 2019-NCOV ACUTE RESPIRATORY DISEASE”, FEBRUARY 4, 2020, BY PETER RICHARDSON ET AL; HTTPS://WWW.THELANCET.COM/JOURNALS/LANCET/ARTICLE/PIIS0140-6736(20)30304-4/FULLTEXT?RSS=YES#ARTICLEINFORMATION
“COVID-19: COMBINING ANTIVIRAL AND ANTI-INFLAMMATORY TREATMENTS”, FEBRUARY 27, 2020, BY JUSTIN STEBBING ET AL; HTTPS://WWW.THELANCET.COM/JOURNALS/LANINF/ARTICLE/PIIS1473-3099(20)30132-8/FULLTEXT
“COVID-19, PROTEIN FOLDING, AND THE CROWD-SOURCED QUEST FOR A CURE”, MARCH 27, 2020, BY LAURA FOSTER, TECHUK; HTTPS://WWW.TECHUK.ORG/INSIGHTS/NEWS/ITEM/17154-COVID-19-PROTEIN-FOLDING-AND-THE-CROWD-SOURCED-QUEST-FOR-A-CURE
“IBM, AMAZON, GOOGLE AND MICROSOFT PARTNER WITH WHITE HOUSE TO PROVIDE COMPUTE RESOURCES FOR COVID-19 RESEARCH”, MARCH 22, 2020, BY FREDERIC LARDINOIS; HTTPS://TECHCRUNCH.COM/2020/03/22/IBM-AMAZON-GOOGLE-AND-MICROSOFT-PARTNER-WITH-WHITE-HOUSE-TO-PROVIDE-COMPUTE-RESOURCES-FOR-COVID-19-RESEARCH/
“INTRODUCING NEW LEVELS OF TRANSPARENCY WITH AI – THOUGHT LEADERS”, APRIL 13, 2020, BY BALAKRISHNA D R AT UNITE.AI; HTTPS://WWW.UNITE.AI/INTRODUCING-NEW-LEVELS-OF-TRANSPARENCY-WITH-AI-THOUGHT-LEADERS/
“LILLY BEGINS CLINICAL TESTING OF THERAPIES FOR COVID-19”, APRIL 10, 2020, BY ELI LILLY PRESS RELEASE; HTTPS://INVESTOR.LILLY.COM/NEWS-RELEASES/NEWS-RELEASE-DETAILS/LILLY-BEGINS-CLINICAL-TESTING-THERAPIES-COVID-19
“NEW AI DIAGNOSTIC CAN PREDICT COVID-19 WITHOUT TESTING”, MAY 11, 2020, BY KING’S COLLEGE LONDON, SCIENCEDAILY; HTTPS://WWW.SCIENCEDAILY.COM/RELEASES/2020/05/200511112628.HTM
“POTENTIAL TREATMENT FOR COVID-19 IDENTIFIED BY BENEVOLENTAI USING ARTIFICIAL INTELLIGENCE ENTERS RANDOMISED CLINICAL TRIAL”, APRIL 10, 2020, BY BENEVOLENTAI PRESS RELEASE; HTTPS://WWW.BENEVOLENT.COM/NEWS/POTENTIAL-TREATMENT-FOR-COVID-19-IDENTIFIED-BY-BENEVOLENTAI-USING-ARTIFICIAL-INTELLIGENCE-ENTERS-RANDOMISED-CLINICAL-TRIAL
“PUTTING AI TO WORK AGAINST COVID-19”, APRIL 21, 2020, BY PHILIPPA BRICE, UNIVERSITY OF CAMBRIDGE, PHG FOUNDATION; HTTPS://WWW.PHGFOUNDATION.ORG/BLOG/PUTTING-AI-TO-WORK-AGAINST-COVID-19
“REAL-TIME TRACKING OF SELF-REPORTED SYMPTOMS TO PREDICT POTENTIAL COVID-19”, MAY 11, 2020, BY CRISTINA MENNI ET AL, NATURE MEDICINE; HTTPS://WWW.NATURE.COM/ARTICLES/S41591-020-0916-2
REUTERS, HTTPS://WWW.REUTERS.COM/ARTICLE/US-IMF-WORLDBANK-OUTLOOK/GLOBAL-ECONOMY-IN-2020-ON-TRACK-FOR-SHARPEST-DOWNTURN-SINCE-1930S-IMF-IDUSKCN21W1MA
“SUCCESS IN CREATING AI WOULD BE THE GREATEST EVENT IN HUMAN HISTORY. UNFORTUNATELY, IT MIGHT BE THE LAST, UNLESS WE LEARN HOW TO AVOID THE RISKS”, MARCH 20, 2020, BY BBSCORP2020, BOSTON BUSINESS SOLUTIONS PVT LTD; HTTPS://BBSCORP.IN/2020/03/30/SUCCESS-IN-CREATING-AI-WOULD-BE-THE-BIGGEST-EVENT-IN-HUMAN-HISTORY-UNFORTUNATELY-IT-MIGHT-ALSO-BE-THE-LAST-UNLESS-WE-LEARN-HOW-TO-AVOID-THE-RISKS/
“TESTING FOR COVID-19”, CDC; HTTPS://WWW.CDC.GOV/CORONAVIRUS/2019-NCOV/SYMPTOMS-TESTING/TESTING.HTML
“THE IMPORTANCE OF DIAGNOSTIC TESTING FOR COVID-19”, BY HEATHER JONES, FUTURE SCIENCE GROUP; HTTPS://WWW.ID-HUB.COM/2020/04/02/THE-IMPORTANCE-OF-DIAGNOSTIC-TESTING-FOR-COVID-19/
“US RESEARCHERS TO ASSESS REMDESIVIR PLUS BARICITINIB FOR COVID-19”, MAY 11, 2020, BY CLINICAL TRIALS ARENA; HTTPS://WWW.CLINICALTRIALSARENA.COM/NEWS/COVID-19-TRIAL-REMDESIVIR-PLUS-BARICITINIB/