Imagine for a moment that a global pandemic breaks out, and suddenly there is a need to find and rapidly develop a cure for a new virus that has emerged worldwide. What can be done? Developing a new drug is a lengthy and expensive process, requiring FDA approval, which takes a long time to obtain. It is desirable to have an alternative and faster pathway for finding a cure for a new virus. In this workshop, we are exploring the field known as Drug Repurposing. Our goal is to find out which existing drugs, already approved by the FDA and used safely by many people for various (but similar) diseases, can also treat the new virus. Discovering such drugs can provide a rapid response to the new virus because they have already received FDA approval.

We continue the research of Dr. Sarel Cohen in collaboration with the University of Potsdam in Germany, who have developed an algorithm based on deep learning for studying a Knowledge Graph containing relationships between viruses, compounds, and genes. Using GNN (Graph Neural Networks), we predict new relationships between drugs and new viruses (in this case, focusing on the coronavirus and its 32 strains), solving a problem called Link Prediction between the nodes representing drugs and those representing the new viruses (strains of coronavirus). We obtain a matrix representing the forecast of interactions between drugs and the coronavirus, and now in the final stage of computation, we perform post-processing to find a list of 100 drugs that can treat the coronavirus.
The innovation in our research lies in the attempt to implement the theoretical algorithm of Prof. Michal Parnas from the field of property testing to check if a set of points is (k,b)-Clusterable and to try to find a practical use for it as part of the post-processing stage, where we attempt to perform optimal clustering.

The research was conducted in collaboration with Prof. Michal Parnas, Mr. Kfir Avles, a master-degree student of Sarel who is researching this topic, and an international research group.