Algorithm accelerated: machine-learning may reduce drug formulation development times

Machine-learning models demonstrate the potential to accelerate drug formulation discovery and development for long-acting injectables.

Long-acting injectables encompass advanced drug delivery systems that release the therapeutic component of the drug over long periods of time, allowing for a prolonged therapeutic effect. Such injectables can increase patient adherence and increase therapeutic efficacy by injecting close to the site of action. A reduction of side effects has also been reported. These advantages have positioned long-acting injectables as one of the most promising modes of administration for chronic diseases. However, one of the challenges associated with long-term injectables involves achieving the optimal amount of drug released over a specified period, requiring drug formulation candidates to be developed, characterised, and authorised through extensive experiments. As a result, bottlenecks within the development of these therapeutics have increased more than other types of drug formulations. 

Professor in pharmaceutical sciences at the University of Toronto Christine Allen commented that “This study takes a critical step towards data-drive drug formulation development with an emphasis on long-acting injectables. We’ve seen how machine-learning has enabled incredible leap-step advances in the discovery of new molecules that have the potential to become medicines. We are now working to apply the same techniques to help us design better drug formulations and, ultimately, better medicines.” 

The team at the University of Toronto trained and evaluated 11 different models of machine-learning tools to accurately predict the rate of drug release. These models included multiple linear regression, random forest, and light gradient boosting machine, among others. Results provided by the models’ test set were compared with previous experimental data, with tree-based models delivering the most accurate results. Further analysis of the data sets were able to extract drug design criteria from the machine-learning models. In particular, the light gradient boosting machine-learning model was utilised to design a long-acting injectable formulation for an ovarian cancer drug currently available. The drug resulting drug release rate was further tested and validated by the light gradient boosting machine model. The formulation produced the slow-release rate sought after by the researchers. Pauric Bannigan, a research associate at the University of Toronto, stated: “This [is] significant because in the past it might have taken us several iterations to get to a release profile that looked like this. With machine-learning, we got there in one.” 

While challenges remain for the complete usage of machine-learning to reduce reliance on trial-and-error testing for the development of long-acting injectables, mainly due to the lack of open-source data available for pharmaceutical sciences, the team hope to work towards this development of robust databases in pharmaceuticals. “For this study, our goal was to lower the barrier of entry to applying machine-learning in pharmaceutical sciences...We’ve made our data sets fully available so others can hopefully build on this work. We want this to be the start of something and not the end of the story for machine-learning in drug formulation,” Bannigan stated. 

Source: University of Toronto scientists use machine | EurekAlert!