Novel Machine Learning Approach Predicts Drug Dissolution Profiles

Strategic Insights -

Research conducted by Chih-Yuan Tseng and colleagues from Sinoveda Canada Inc. has developed a novel machine learning approach to predict drug dissolution profiles in the gastrointestinal tract. The study, published in the Journal of Chemical Information and Modeling, presents a two-stage machine learning approach that integrates physics-informed neural networks (PINNs) and deep neural networks (DNNs) to predict dissolution profiles in water with varying concentrations of surfactant Sodium Lauryl Sulfate.

This approach shows significant potential as a low-cost, time-efficient tool for early phase drug formulation, with an average testing accuracy of 61.7% at an 80:20 train-to-test split. Although the current accuracy is below the generally acceptable range of 70-80%, the researchers expect future improvements as data quality and diversity increase.

Key Takeaways:

  • The research presents a novel two-stage machine learning approach, combining PINNs and DNNs to predict drug dissolution profiles.
  • The PINNs stage extracts key dissolution parameters, such as the dissolution rate constant (k) and the dissolved mass fraction at saturation (ph), from existing dissolution data.
  • The extracted parameters are then used to train a DNN to predict dissolution profiles based on the drug's chemical structure and dissolution medium.
  • The DNN with 128 neurons in two hidden layers and a learning rate of 0.1 achieved an average testing accuracy of 61.7%.
  • The research has been peer-reviewed and published in the Journal of Chemical Information and Modeling.
  • The approach shows significant potential for early phase drug formulation, but requires further improvement in data quality and diversity.

Statistics:

  • The PINNs stage used 8 hidden layers with 40 neurons per layer.
  • The DNN used 2 hidden layers with 128 neurons per layer and a learning rate of 0.1.
  • The approach reported an average testing accuracy of 61.7%.
  • The research highlights the need for further improvement in data quality and diversity to achieve acceptable accuracy.

Sources:

  • NewsRx. Data on Machine Learning Reported by Chih-Yuan Tseng and Colleagues (Machine Learning Based Quantitative Structure-Dissolution Profile Relationship). Journal of Engineering. June 16, 2025; p 459.