Nikolaos Sidiropoulos, Louis T. Rader Professor of electrical and computer engineering at the University of Virginia, and his collaborators are developing novel data mining and artificial intelligence methods to help in the fight against COVID-19.
Sidiropoulos and his Ph.D. student Charilaos I. Kanatsoulis, a post-doctoral researcher at the University of Pennsylvania, leveraged their expertise in coupled matrix and tensor factorization to speed the search for existing drugs that can be repurposed to treat the novel coronavirus. Working with a publicly released data set encompassing 8,103 drugs targeting many different diseases, and pairwise and three-way interactions between genes, diseases, symptoms, drugs, molecular pathways, and other entities, Sidiropoulos and Kanatsoulis came up with a method that produced a list of top-100 drugs most likely to help in the fight against COVID-19. Their method offers 100% better precision over the prior art, retrieving one-third of the drugs used in clinical trials for COVID-19 and putting Dexamethasone at the top of their top-100 list, consistent with the latest clinical results. The method developed by Sidiropoulos and Kanatsoulis is purely association data-driven; it does not use any human domain-expert knowledge. Sidiropoulos and Kanatsoulis will present their paper, TeX-Graph: Coupled tensor-matrix knowledge-graph embedding for COVID-19 drug repurposing, at the SIAM International Conference on Data Mining, April 29-May 1.