SCI和EI收录∣中国化工学会会刊

中国化学工程学报 ›› 2024, Vol. 66 ›› Issue (2): 263-272.DOI: 10.1016/j.cjche.2023.09.006

• Full Length Article • 上一篇    下一篇

Machine learning with active pharmaceutical ingredient/polymer interaction mechanism: Prediction for complex phase behaviors of pharmaceuticals and formulations

Kai Ge, Yiping Huang, Yuanhui Ji   

  1. Jiangsu Province Hi-Tech Key Laboratory for Biomedical Research, School of Chemistry and Chemical Engineering, Southeast University, Nanjing 211189, China
  • 收稿日期:2023-06-08 修回日期:2023-09-17 出版日期:2024-02-28 发布日期:2024-04-20
  • 通讯作者: Yuanhui Ji,E-mail:yuanhuijinj@163.com
  • 基金资助:
    The authors would like to acknowledge the financial support from the National Natural Science Foundation of China (22278070, 21978047, 21776046).

Machine learning with active pharmaceutical ingredient/polymer interaction mechanism: Prediction for complex phase behaviors of pharmaceuticals and formulations

Kai Ge, Yiping Huang, Yuanhui Ji   

  1. Jiangsu Province Hi-Tech Key Laboratory for Biomedical Research, School of Chemistry and Chemical Engineering, Southeast University, Nanjing 211189, China
  • Received:2023-06-08 Revised:2023-09-17 Online:2024-02-28 Published:2024-04-20
  • Contact: Yuanhui Ji,E-mail:yuanhuijinj@163.com
  • Supported by:
    The authors would like to acknowledge the financial support from the National Natural Science Foundation of China (22278070, 21978047, 21776046).

摘要: The high throughput prediction of the thermodynamic phase behavior of active pharmaceutical ingredients (APIs) with pharmaceutically relevant excipients remains a major scientific challenge in the screening of pharmaceutical formulations. In this work, a developed machine-learning model efficiently predicts the solubility of APIs in polymers by learning the phase equilibrium principle and using a few molecular descriptors. Under the few-shot learning framework, thermodynamic theory (perturbed-chain statistical associating fluid theory) was used for data augmentation, and computational chemistry was applied for molecular descriptors’ screening. The results showed that the developed machine-learning model can predict the API-polymer phase diagram accurately, broaden the solubility data of APIs in polymers, and reproduce the relationship between API solubility and the interaction mechanisms between API and polymer successfully, which provided efficient guidance for the development of pharmaceutical formulations.

关键词: Multi-task machine learning, Density functional theory, Hydrogen bond interaction, Miscibility, Solubility

Abstract: The high throughput prediction of the thermodynamic phase behavior of active pharmaceutical ingredients (APIs) with pharmaceutically relevant excipients remains a major scientific challenge in the screening of pharmaceutical formulations. In this work, a developed machine-learning model efficiently predicts the solubility of APIs in polymers by learning the phase equilibrium principle and using a few molecular descriptors. Under the few-shot learning framework, thermodynamic theory (perturbed-chain statistical associating fluid theory) was used for data augmentation, and computational chemistry was applied for molecular descriptors’ screening. The results showed that the developed machine-learning model can predict the API-polymer phase diagram accurately, broaden the solubility data of APIs in polymers, and reproduce the relationship between API solubility and the interaction mechanisms between API and polymer successfully, which provided efficient guidance for the development of pharmaceutical formulations.

Key words: Multi-task machine learning, Density functional theory, Hydrogen bond interaction, Miscibility, Solubility