Corona-AI UK: Harnessing the power of AI and mobile supercomputing for drug repurposing and food tailoring in the fight against the coronavirus epidemic
As part of the DreamLab: Corona-AI project, a team led by Dr Kirill Veselkov from the Department of Surgery and Cancer at Imperial College London in collaboration with the Vodafone Foundation, are pairing bespoke AI technologies, mobile supercomputing, and big “-omics” data to narrow the search of combinations of existing drugs and food-based drug-like molecules to help in the fight against the coronavirus pandemic.
The ongoing DreamLab crowd computing project has already been using the processing power of thousands of idle smartphones to help uncover anti-cancer properties of everyday foods and medicines (1). Now that processing power is being shifted to help to combat the coronavirus.
Head of the Department of Surgery and Cancer, Professor George Hanna, highlighted the importance of the project:
The rapid spread of the acute respiratory disease (COVID-19) caused by the novel coronavirus has had a massive detrimental impact on human health and the global economy leading to the immediate need for medical and nutritional interventions to help combat the outbreak.
Finding new indications or alternative uses of existing drugs (known as “drug repositioning”) is an effective way to get around the slow and costly process of developing specific medicines to treat this infectious disease. In the current pandemic, it could potentially save thousands of lives. The human diet is rich with drug-like molecules that have been shown to play a role in both the prevention and treatment of viral diseases, by interacting with drugs, or by acting as “medicines” themselves. However, the traditional practicalities involved in investigating the influence of a single drug or food component would take too long to have an impact on this crisis.
At present, the landscape of potential drug-like molecules in food is unimaginably vast. The traditional experimental methodology of investigating the influence of a single drug or food component on any particular viral infection takes months or even years.
The DreamLab: Corona-AI project takes a radically different approach to traditional testing methods. Powered by the general public, the project combines artificial intelligence and the processing power of idle smartphones to speed up the discovery of novel anti-viral components in existing medicines and help the hunt for anti-viral molecules in foods.
The DreamLab team member Professor Vasilis Vasiliou of Yale School of Public Health further emphasized the importance of the project: “In these difficult times AI can do a world of good in the race against Covid-19. The Dreamlab team is utilizing the power of AI to screen approved drugs that may have the potential to fight Covid-19 and also to identify food components that can boost our immunity against the effects of the virus”.
Innovative network-driven AI for drug repositioning and discovery of anti-viral molecules in foods
Coronaviruses cannot survive or replicate without the help of their hosts. In fact, all viruses have naturally evolved a sophisticated arsenal of molecular strategies to exploit the host’s cellular machinery for their own benefit of survival and replication. These strategies rely on a complex network of physical interactions between viral and host proteins, the so-called ‘virus-host interactome networks’ (2).
The traditional anti-viral drug development paradigm is ‘one drug for one viral protein target’. This approach has multiple drawbacks, among which is the mutation of the virus that can quickly make the drug inefficient or useless. Instead, we need to target a whole virus-host interactome. Even though exciting progress has been made, the preventive vaccines and drug effects against SARS-Cov-2 specific protein targets are likely to be beset by the generation of viral escape mutants.
Taking advantage of our previous work in cancer, the Corona-AI DreamLab project “aims to build up a clearer picture of which individual or combinations of molecules are most suited to disrupt molecular host-viral interactome networks essential for the coronavirus survival, not just its specific proteins”, according to Dr. Kirill Veselkov. These molecules could be either existing drugs (i.e. drugs not known previously or used for anti-viral treatments) or drug-like molecules within foods. The outcomes can potentially shed light on multi-drug therapies alongside dietary interventions against disrupted host-viral interactome networks in humans.
A member of the DreamLab collaboration, Professor Michael Bronstein, Chair in Machine Learning and Pattern Recognition at the Department of Computing Imperial College London and Head of Graph Learning Research at Twitter, described the algorithmic core of the project:
Combinatorial nature of the problem: Why DreamLab?
The progress of this research was originally slowed by a lack of access to supercomputing. Combinations of three, four, or even more compounds would be impossible to test in the lab. Let's say there are 10,000 molecules, with different combinations - that's a trillion possibilities that need to be computationally “tested” against virus-host interactomes of coronavirus strains, something that would be unmanageable on a normal computer. However, the DreamLab App uses machine learning on a mobile supercomputing network, to analyse billions of combinations of existing drug, food-based molecules and genetic interactions, fundamentally reducing the time needed to make discoveries.
Member of the project team, Dr Reza Mirnezami, Consultant Colorectal Surgeon, Royal Free Hospital, said: “The DreamLab team are working on harnessing the power of AI to identify how commonly used, approved drugs could be ‘redeployed’ in the war against COVID-19. Moreover, we are looking at how to improve outcomes in COVID-19 patients using diet, which will undoubtedly affect host immunity and gut microbiotal resilience.”
During this initial phase of the Corona-AI project, we aim to test combinations of up to two molecules of existing drugs or drug-like molecules in foods against host-virus interactomes of coronavirus strains including SARS-Cov-2 - the causative agent responsible for the current COVID-19 pandemic. The DreamLab smartphone network of 100K users has the combined power of a supercomputer that can crunch the data from this phase in a time that otherwise would require decades using a standard desktop computer.
Download the DreamLab now to help us fight COVID-19 together.
Author: Dr Kirill Veselkov; Lecturer in Computational Medicine and Cancer Informatics, Department of Surgery and Cancer, Imperial College London; Assistant Professor Adjunct of Epidemiology, Yale School of Public Health, USA.
Veselkov K, Gonzalez G, Shahad A, Galea D, Mirnezami R, Youssef J, Bronstein M & Laponogov I. HyperFoods: Machine-intelligent searching for cancer-beating molecules in foods. Nature Scientific Reports, 2019, 1-11
de Chassey, B., Meyniel-Schicklin ,L., Vonderscher,J., Andre,P. and Lotteau,V. (2014) Virus-host interactomics: new insights and opportunities for antiviral drug discovery. Genome Medicine. 6 ,115