Researchers Predict Drug Interactions with Machine Learning to Enhance Patient Safety

Researchers from the Massachusetts Institute of Technology (MIT), Brigham and Women’s Hospital, and Duke University have devised a novel AI-powered strategy to pinpoint the transporters utilized by different drugs as they traverse the digestive tract. This innovative approach has already unveiled potential interactions between a commonly prescribed antibiotic and a blood thinner, highlighting the critical role of transporter proteins in drug absorption.

Related article: Absci Accelerates Breakthroughs in AI-Designed Drugs with AstraZeneca and Almirall Collaborations

Identifying Drug Transporters and Clarifying Their Roles

Transporter proteins located on the cells lining the gastrointestinal (GI) tract play a crucial role in drug absorption of orally administered drugs through the digestive system. Identifying the transporters used by specific drugs to leave the GI tract could be highly beneficial in improving patient treatment because if two drugs rely on the same transporter, they can interfere with each other and should not be prescribed together.

The study, published in Nature Biomedical Engineering on February 20, is led by Dr. Giovanni Traverso, an Associate Professor in MIT’s Department of Mechanical Engineering along with former MIT postdocs Yunhua Shi and Daniel Reker, and it examines the interaction between drugs and intestinal transporters.

The team has chosen three commonly used transporter proteins in the GI tract, namely, BCRP, MRP2, and PgP as the focus of their new study, and employed a multipronged strategy combining a tissue model developed in 2020 and machine-learning algorithms. By systematically exposing laboratory-grown pig intestinal tissue to various drug formulations, the researchers were able to measure drug absorbability and investigate the role of individual transporters. 

Using short RNA strands called “small interfering RNA (siRNA)”, researchers selectively suppressed the expression of key transporters to study their interactions with different drugs. Through this experimental setup, the team evaluated 23 commonly used drugs and trained a machine-learning model to predict interactions based on chemical similarities between drugs. 

Predicting Drug Interactions with Machine Learning Algorithms

Their analysis encompassed 28 currently used drugs and 1,595 experimental drugs, generating nearly 2 million predictions of potential drug interactions. The machine-learning model, trained on experimental data and drug databases, successfully predicted potential interactions between drugs and transporters. Notably, the model identified an interaction between doxycycline, an antibiotic, and warfarin, a widely prescribed blood thinner. Other drugs predicted to interact with doxycycline include tacrolimus, an immunosuppressant, digoxin, a drug for treating atrial fibrillation and heart failure, and levetiracetam, an anti-seizure medication.

Subsequent analysis using data from about 50 patients at Massachusetts General Hospital and Brigham and Women’s Hospital confirmed the model’s predictions. In one example, when patients taking warfarin were given doxycycline, their blood levels of warfarin increased accordingly, before dropping back down after stopping the antibiotic treatment. Also, these real-world data confirmed the model’s predictions regarding the effect on the absorption of doxycycline digoxin, levetiracetam, and tacrolimus. Among them, only tacrolimus had been previously suspected to interact with the antibiotic.

“We are the first to predict this interaction using this accelerated in silico and in vitro model,” commented Prof. Traverso. “This kind of approach gives you the ability to understand the potential safety implications of giving these drugs together.”

AI Holds Promise for Enhancing Drug Safety and Development

Apart from the funding by MIT and Brigham and Women’s Hospital, this groundbreaking has also received financial support from the U.S. National Institutes of Health. The AI-powered approach not only offers insights into existing drug combinations but also holds promise for optimizing the development of new drugs. By understanding drug transport mechanisms, drug developers can tailor formulations of new drug molecules to prevent undesirable interactions and enhance their absorbability, paving the way for safer and more effective treatments.

Notably, Vivtex, a biotech company co-founded in 2018 by former MIT postdoc Thomas von Erlach (currently the Chief Scientific Officer), MIT Institute Professor Robert Langer, and Prof. Traverso, is actively pursuing strategies to leverage the above-mentioned AI-based technology in oral drug delivery systems. In January 2024, Vivtex signed a research collaboration agreement with Astellas Pharma, which will use the company’s unique and proprietary screening and formulation platform technology to support the development of a novel oral drug candidate provided by Astellas.

Author

Richard Chau