Study Shows How AI can be Used for Drug Repurposing

A new machine-learning method developed by researchers analyzes huge amounts of data to help establish which existing medications could enhance outcomes in diseases for which they are not prescribed.

The purpose of this study is to expedite drug repurposing, which is not a new idea. Earlier, Botox injections were approved to treat crossed eyes and are currently used as a migraine treatment and top cosmetic strategy to decrease wrinkles.

However, achieving such new uses normally involves a combination of serendipity and time-consuming, high-priced randomized clinical trials to guarantee that a drug deemed effective for one disorder will be beneficial as a treatment for some other disorder.

Scientists from Ohio State University have developed a framework that integrates huge patient care-related datasets with high-powered computation to realize repurposed drug candidates and the evaluated impacts of those existing medications on a specified set of outcomes.

Although the focus of this study was on the proposed repurposing of drugs to avoid stroke and heart failure in patients suffering from coronary artery disease, the framework is flexible—and could even be used for a majority of the diseases.

This work shows how artificial intelligence can be used to ‘test’ a drug on a patient, and speed up hypothesis generation and potentially speed up a clinical trial. But we will never replace the physician—drug decisions will always be made by clinicians.

Ping Zhang, Study Senior Author and Assistant Professor of Computer Science and Engineering and Biomedical Informatics, Ohio State University

The study was published in the journal Nature Machine Intelligence on January 4^th, 2021.

Repurposing of drugs is an appealing quest since it could reduce the risk related to the safety testing of new medications and considerably decrease the time taken to get a drug to the market for clinical use.

Although randomized clinical trials are considered to be the gold standard to find the effectiveness of a drug against a disease, Zhang observed that machine learning can account for hundreds—or even thousands—of human differences in a huge population that could impact how medicine functions in the body.

Such factors, or confounders, varying from age, sex, and race to disease severity and the presence of other illnesses, act as parameters in the deep learning computer algorithm based on which the framework has been developed.

That data is gathered from “real-world evidence,” or longitudinal observational information regarding millions of patients captured by insurance claims or electronic medical records and prescription data.

“Real-world data has so many confounders. This is the reason we have to introduce the deep learning algorithm, which can handle multiple parameters,” stated Zhang, head of the Artificial Intelligence in Medicine Lab and a core faculty member in the Translational Data Analytics Institute at Ohio State.

“If we have hundreds or thousands of confounders, no human being can work with that. So we have to use artificial intelligence to solve the problem. We are the first team to introduce use of the deep learning algorithm to handle the real-world data, control for multiple confounders, and emulate clinical trials,” he added.

The researchers made use of insurance claims data related to almost 1.2 million heart-disease patients, which offered information on their assigned treatment, disease outcomes, and several values for possible confounders.

The deep learning algorithm also has the power to consider the passage of time in each patient’s experience—for every visit, diagnostic test, and prescription. The model input for drugs depends on their active ingredients.

The team employed the so-called causal inference theory to categorize, for this analysis, the active drug and placebo patient groups that form part of a clinical trial. The model monitored patients for two years—and compared the status of their disease at that endpoint to know whether they took medications, the drugs they took, and when they began the regimen.

With causal inference, we can address the problem of having multiple treatments. We don’t answer whether drug A or drug B works for this disease or not, but figure out which treatment will have the better performance.

Ping Zhang, Study Senior Author and Assistant Professor of Computer Science and Engineering and Biomedical Informatics, Ohio State University

They propose that the model would determine drugs that could reduce the risk of stroke and heart failure in coronary artery disease patients.

The model yielded nine drugs that were regarded as likely to offer those therapeutic benefits, three of which are in use at present. This implies that six candidates were determined by the analysis for drug repurposing.

Amid other findings, the analysis indicated that escitalopram, which is used to treat anxiety and depression, and metformin, a diabetes medication, could reduce the risk for stroke and heart failure in the model patient population. Apparently, both of these drugs are presently being tested for their effectiveness against heart disease.

Zhang emphasized that what was discovered by the research team in this case study is less crucial compared to how they arrived at the results.

My motivation is applying this, along with other experts, to find drugs for diseases without any current treatment. This is very flexible, and we can adjust case-by-case. The general model could be applied to any disease if you can define the disease outcome.

Ping Zhang, Study Senior Author and Assistant Professor of Computer Science and Engineering and Biomedical Informatics, Ohio State University

The study was financially supported by the National Center for Advancing Translational Sciences, which funds the Center for Clinical and Translational Science at Ohio State.

Graduate student Ruoqi Liu and research assistant professor Lai Wei from Ohio State also contributed to the study.

Journal Reference:

Liu, R., et al. (2021) A deep learning framework for drug repurposing via emulating clinical trials on real-world patient data. Nature Machine Intelligence. doi.org/10.1038/s42256-020-00276-w.