Scientists are using machine learning to speed the development of drug formulations

University of Toronto scientists are using machine learning to speed drug formulation development
[Left to Right] Kristen Allen and Alan Aspuru-Guzik of the University of Toronto combine expertise in pharmaceutical sciences, artificial intelligence, and machine learning to develop new drug formulations faster. Credit: Steve Southon

Scientists at the University of Toronto have successfully tested the use of machine learning models to guide the design of long-acting injectable drug formulations. The potential for machine learning algorithms to speed drug formulation could reduce the time and cost associated with drug development, making promising new drugs available faster.

The study was published today in Nature Communications He is one of the first applicants machine learning Techniques for designing injectable long-acting polymeric drug formulations.

The interdisciplinary research was led by Christine Allen of the University of Toronto’s Department Pharmaceutical sciences and Alán Aspuru-Guzik, from the Departments of Chemistry and Computer Science. Both researchers are also members of the Acceleration Consortium, a global initiative using Artificial intelligence and automation to speed discovery of materials and molecules needed for sustainable future.

“This study takes a critical step toward advancing data-driven drug formulation with a focus on long-acting injectables,” said Christine Allen, associate professor of pharmaceutical sciences at the University of Toronto’s Leslie Dunn School of Pharmacy. “We’ve seen how machine learning has enabled amazing advances in discovering new molecules that have the potential to become medicines. We’re now working on applying the same techniques to help us design better drug formulations and, ultimately, better medicines.”

One of the most promising therapeutic strategies for treating chronic diseases is long-acting injectables (LAI), a class of advanced drug delivery systems designed to release their payload over extended periods of time to achieve a long-term therapeutic effect. This approach could help patients better adhere to their medication regimen, reduce side effects, and increase efficacy when injected close to the site of impact in the body. However, achieving the optimal amount of drug release over the desired time period requires the development and characterization of a wide range of formulation candidates through extensive and time-consuming trials. This trial-and-error approach has created a significant bottleneck in the development of LAI compared to traditional types of drug formulation.

He said “AI is changing the way we use science. It helps accelerate discovery and improvement. This is an excellent example of a ‘pre-AI’ and a ‘post-AI’ moment and shows how drug delivery can be affected by this interdisciplinary research.” Alán Aspuru-Guzik, Professor of Chemistry and Computer Science at the University of Toronto who also holds the CIFAR Chair for Artificial Intelligence Research at the Vector Institute in Toronto.

To investigate whether machine learning tools can accurately predict drug release rate, the research team trained and evaluated a series of eleven different models, including multiple linear regression (MLR), random forest (RF), and light gradient enhancement machine (lightGBM). , and neural networks (NN). The dataset used to train a selection of machine learning models was generated from previously published studies by the authors and other research groups.

“Once we had the dataset, we divided it into two subsets: one used to train the models and one to test. We then asked the models to predict the results of the test set and directly compared it to previous experimental data. We found that For Pharmacy, University of Toronto:

As a next step, the team worked to apply these predictions and demonstrate how machine learning models can be used to inform the design of new LAIs. The team used advanced analytical techniques to extract design criteria from lightGBM. Model. This allowed the design of a new LAI formulation for a drug currently used to treat ovarian cancer. “Once you have a trained model, you can then work on interpreting what the machine has learned and use that to develop design standards for new systems,” Bannigan said. Once prepared, the drug release rate was tested and the predictions made by the LightGBM model were validated. “The formula definitely had the slow release rate we were looking for. This was important because in the past it would have taken us many iterations to get to a release profile that looked something like this, with machine learning we got it right,” he said.

The results of the current study are encouraging and point to the potential for machine learning to reduce reliance on trial-and-error testing slowing the pace of development of long-acting injectables. However, the study authors specify that the lack of available open source data sets in the pharmaceutical sciences presents a major challenge to future progress. “When we began this project, we were surprised by the lack of data reported across the many studies using polymeric microparticles,” Allen said. “This means that the studies and work that have been done to develop the machine learning models that we need to drive progress in this area cannot be leveraged,” Allen said. “There is a real need to create robust databases in the pharmaceutical sciences open access and available to all so that we can work together to advance this field.”

To advance the move toward the accessible databases needed to support the integration of machine learning into the pharmaceutical sciences more broadly, Allen and the research team made their work datasets And the blade It is available on the Zenodo open source platform.

“For this study, our goal was to reduce the barrier to entry for the application of machine learning in the pharmaceutical sciences,” said Banigan. “We’ve made our datasets fully available so others can build on this work. And we want this to be the beginning of something, not the end of the machine learning story in drugs editing “.

more information:
machine learning models to accelerate the design of long-acting polymeric injections, Nature Communications (2023). DOI: 10.1038/s41467-022-35343-w

Introduction of
University of Toronto


the quote: Scientists Use Machine Learning to Accelerate Drug Formulation Development (2023, January 10) Retrieved January 10, 2023 from https://phys.org/news/2023-01-scientists-machine-fast-track-drug.html

This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without written permission. The content is provided for informational purposes only.

Leave a Comment