AI finds the effects of other diseases of existing drugs

When developing a new drug, it takes a long time because it is necessary to obtain approval from the authorities for safety and effectiveness by conducting experiments in several stages of cell testing, animal testing and clinical testing. Therefore, it is possible to shorten the time by converting drugs that have already been recognized for efficacy and safety against diseases to other diseases, but a clinical trial that takes time is still necessary. To save human resources and time for these drugs, researchers devised a method of estimating drug candidates and effects in an AI algorithm that learned large amounts of data.

Relocating drugs using drugs that are already effective against diseases to treat other diseases is not a new concept. Botox, which is used for the treatment of strabismus, is typically used in the field of cosmetic surgery, and antiparasitic drugs are considered promising as a treatment for COVID-19 infection, for example.

Since drug recycling uses drugs that have already been approved for human safety, it has the advantage of shortening the time until approval as a treatment rather than developing a new drug. Still, it’s a task that takes a lot of time and talent to prove that it works against the disease through randomized clinical trials.

To tackle this problem, Ohio State University researchers have developed a framework that uses large-scale patient datasets and AI to speculate on promising candidates for revich and the possible effects. The study looked at relocations to prevent heart failure and stroke in patients with arterial disease, but the framework itself could be used for many other disease relocations.

According to the research team, this study is the first to emulate a clinical trial by processing real data using deep learning algorithms and adjusting several confounding factors. Data used in computing is collected from millions of patients in the real world, such as electronic medical records, insurance claim information, and prescriptions. Real data has many confounding factors, so it is necessary to use deep learning algorithms that can handle multiple parameters. Hundreds to thousands of confounding factors are difficult for humans to handle and AI is needed to solve problems. Related information can be found here .