The machine learning method can crush data to find new uses for existing drugs

Scientists have developed a machine learning method that compresses a lot of data to help determine which existing drugs may produce results in over-the-counter diseases.

The purpose of this work is to accelerate the reproduction of drugs, which is not a new concept – think of Botox injections, which were originally agreed to treat cross-eyes and now the treatment of migraine and high cosmic strategy to reduce the appearance of wrinkles.

But access to these new practices usually involves a combination of randomized and time-consuming and expensive clinical trials to ensure that a drug that is considered effective for one disorder is useful as a treatment for something else. .

Ohio State University researchers created a framework that combines big data related to patient care with high-powered computing to reach regenerative drug candidates and the estimated effects of these existing drugs on a set of identified outcomes.

While this study focused on drug replacement recommended for the prevention of heart failure and stroke in patients with coronary artery disease, the framework is flexible – and could be applied. involved in most diseases.

This work demonstrates how artificial intelligence can be used to diagnose a ‘drug’ on a patient, and to accelerate hypothesis generation and potentially accelerate clinical trial. But we will never replace the doctor – drug decisions are always made by clinicians. “

Ping Zhang, Lead Author, Associate Professor of Computer Science and Engineering and Biomedical Informatics, Ohio State

The research is published today (Jan. 4, 2021) in Nature machine knowledge.

Reproduction of drugs is an attractive pastime as it may reduce the risk associated with the safety testing of new medicines and significantly reduce the time it takes to get a drug into the market for clinical use.

Randomized clinical trials are the gold standard for determining the effectiveness of a drug against disease, but Zhang noted that machine learning can account for hundreds – or thousands – of human side effects. within a large population that can affect how medicine works in the body. These factors, or antagonists, range from age, sex and race to the depth of disease and the presence of other diseases, acting as parameters in the deep learning computer algorithm on which the framework is based.

That information comes from “real-world evidence,” which is the long-term observation data of millions of patients captured by electronic medical records or insurance claims and drug data.

“Real-world data has so many challenges. This is why we need to introduce the deep learning algorithm, which can handle several parameters,” said Zhang, who directs the Artificial Intelligence in Medicine Lab and is a founding faculty member in Translation. Ohio State Institute of Data Analysis. “If we have hundreds or thousands of opponents, no one can work with that. So we have to use artificial intelligence to solve the problem.

“We are the first team to introduce the use of the deep learning algorithm to manipulate real-world data, control for multitasking, and report on clinical trials,” Zhang said.

The research team used insurance claims data on nearly 1.2 million patients with heart disease, which provided information about their prescribed treatment, disease outcomes and different values ​​for potential antidepressants. The deep learning algorithm also has the power to pay attention to the movement of time in each patient’s experience – for each visit, prescription and diagnostic test. The model input for drugs is based on their active ingredients.

Applying what is known as causal decision theory, the researchers classified, for the purposes of this analysis, the active drug and placebo patient groups that would be detected in a clinical trial. The model monitored patients for two years – and compared their disease status at that final stage in terms of whether they took medications, what drugs they took and when they started the regimen.

“With causal decision, we can address the problem of multiple medications. We will not answer whether drug A or drug B works for this disease or not, but we do get find out which treatment will improve performance, “Zhang said.

Their hypothesis: that the model would identify drugs that may reduce the risk for heart failure and stroke in patients with coronary artery disease.

The model identified nine drugs that were thought to provide these therapeutic benefits, three of which are currently in use – meaning that the analysis identified six candidates for drug replacement. Among other findings, the analysis suggested that diabetes medication, metformin, and escitalopram, used to treat depression and anxiety, may lower the risk for heart failure and stroke in a population. patients. As it turns out, both of these drugs are currently being tested for their effectiveness against heart disease.

Zhang confirmed that the team ‘s findings in this case study are not as important as how they got there.

“My motivation is applying this, along with other experts, to find drugs for diseases without any conventional treatment. This is very flexible, and we can change case-by-case,” he said. “The general model could be applied to any disease if you can explain the outcome of the disease.”