Virtual discussion assesses role of machine learning and data science in developing quicker and more targeted treatments – including as a response to COVID-19
Using Artificial Intelligence to help repurpose existing drugs so they treat new illnesses has a “bright future”, but the world needs to be more prepared to share health data, a virtual panel hosted by Qatar Foundation’s global health initiative has been told.
The World Innovation Summit for Health (WISH) brought together experts from the fields of biomedicine, computing, and healthcare collaboration, as well as the World Health Organization (WHO), to explore how AI and bioinformatics could support the modification of already-approved medicines to provide more rapid solutions for the likes of COVID-19.
The event formed part of Qatar Foundation’s contribution to Global Goals Week, during which it has staged a series of sessions themed around developing local solutions to worldwide challenges. Among those who joined the WISH-organized discussion were scientists from Qatar Computing Research Institute (QCRI) and Qatar Biomedical Research Institute (QBRI), both part of QF member Hamad Bin Khalifa University.
Speaking at Calculating The Cure: AI and COVID-19, Dr. Raghvendra Mall, a scientist at QCRI, said: “If you have 20,000 approved drugs on the market and want to test each of them in a clinical setting for a set of patients, it would cost billions of dollars and tens of thousands of hours, and it’s almost unrealistic to do that.
The goal is to find which set of drugs can be tried first and that are most likely to be effective, and this is where different machine learning models can be used to reach a consensus
“AI models can help, such as through screening and helping to prioritize which of these 20,000 drugs will potentially be most suitable in blocking diseases such as COVID-19, and this is what we are trying to do. The goal is to find which set of drugs can be tried first and that are most likely to be effective, and this is where different machine learning models can be used to reach a consensus.
“The costs and the time associated with bringing an approved drug onto the market [for a different treatment] are almost half of that required for a novel compound. There have been many success stories in drug repurposing, especially in oncology and cancer, and while AI-based drug repurposing is relatively new, there is a plethora of researchers and biotech companies trying to integrate multiple layers of information and pass them to machine learning and deep learning-based modes, so they can repurpose drugs to treat other diseases.
“With the availability of a large amount of patient and clinical trial data, this can be processed by machine learning-based models that act as a support system to help policymakers, healthcare providers, and society as a whole, so I feel the future is bright for AI-based drug repurposing.”
It would be ignorant not to take advantage of AI because it allows us to point ourselves in the right direction, and we are now able to carry out this work at a biological level
Dr. Prasanna Kolatkar, Senior Scientist at QBRI, told the webinar that the myriad ways that proteins – “the active players and engines for what makes us” – interact within the body means biological data science cannot yet be 100 percent accurate, but added: “It’s clearly helpful and is changing the way science works – it would be ignorant not to take advantage of AI because it allows us to point ourselves in the right direction, and we are now able to carry out this work at a biological level.
“People may say that a repurposed drug is already proven to be safe. However, by understanding it as a molecular level, you can have a good idea of what is binding and what the possibility of off-target drug effects is, and even modify it in the future to make even better drugs. It can definitely save time and, depending on how severe the disease is, perhaps it will eventually be possible to allow certain repurposed drugs to be released earlier if something is needed rapidly to treat a disease among certain groups.”
We also need to protect the privacy and confidentiality of data to bring trust to the ecosystem, for us to be able to reap all the benefits of data and digital technologies
Also joining the discussion, moderated by WISH’s Head of Content Maha El Akoum, was Bernardo Mariano Junior, Chief Information Officer at WHO, who said health data is one of the major gaps in the global health sector, and a governance framework that allows data scientists to help speed the response to diseases – which WHO is currently working on this as part of a global digital health strategy – is needed.
“The tech industry uses data to monetize it, but the health sector has a culture of donation rather than monetization,” he said. “We need to bring these two worlds together and really ensure we leverage data for global good and to have a positive impact on health.
“WHO is the custodian of data from its 194 member states, so the question is whether countries would be willing to share data with us for global good and ensure an AI-supported device can mine it to deliver better results and learn more through machine learning algorithms. Alongside helping data scientists, we also need to protect the privacy and confidentiality of data to bring trust to the ecosystem, for us to be able to reap all the benefits of data and digital technologies.”
Machine learning approaches with prediction models could be quite important if we have to decide where best to target resources, such as when prioritizing the rollout of a COVID-19 vaccine
Meanwhile, Professor Aziz Sheikh, Director at the University of Edinburgh’s Usher Institute – which works with populations and data to advance public health through innovative collaborations – said: “We’re still at a pretty early stage of the journey, but we’re already beginning to see a number of ways in which the field of AI can interact with public health provision – diagnosis, trying to identify potential cures and treatments and see which ones may get traction, and to guide decision-making by clinicians and increasingly by patients.
“Machine learning approaches with prediction models could be quite important if we have to decide where best to target resources, such as when prioritizing the rollout of a COVID-19 vaccine. We are talking about the potential vaccination of four billion-plus people and nothing like that has ever been done before. There will be capacity issues, so an approach like this could be useful in prioritizing those at risk of the most severe outcomes.”
Professor Sheikh, who is the lead author of the WISH 2018 report on Harnessing Data Science and AI in Healthcare, also highlighted the use of AI to study public sentiment, issues, and misapprehensions about COVID-19 responses and policies; estimate the impact of measures such as lockdowns and face masks and the “changing shape” of the pandemic’; and increasing the use of robots for tasks such as sanitizing healthcare environments.
The virtual WISH 2020 Summit will take place from November 15-19. To register, visit www.wish.org.qa