Screening of bioactive compounds and deep learning-driven quality control of Angong Niuhuang pills

Angong Niuhuang Pills (AGNHP), a famous Chinese medicine compound preparation, is widely used to treat stroke and other brain disorders. However, owing to the complexity of herbal components and diversity of production processes, quality control is challenging.

This study aimed to explore deep learning-driven quality control methods to accurately and efficiently determine the content of key active compounds in AGNHP and to realize large-scale quality control and yield monitoring.

In this study, AGNHP was used to screen and validate active compounds using liquid chromatography-mass spectrometry (LC-MS) combined with network pharmacology and partial least squares (PLS) analysis. The anti-inflammatory activity was validated in zebrafish and cellular models, and quality control was achieved using near-infrared spectroscopy (NIR) and the deep learning model. This study aimed to establish a comprehensive quality-control system for AGNHP.

Seven key active compounds, including taurocholic acid, and chenodeoxycholic acid, were screened using LC-MS analysis and network pharmacological prediction. These compounds showed significant anti-inflammatory activity in vitro and in vivo models. BiGRU-MAR model was able to accurately predict the levels of these compounds, which were in high agreement with the actual measured values, demonstrating its effectiveness in AGNHP quality control.

This study established a complete quality control system for the AGNHP and promoted its scientific and standardized quality analysis and control. The results of this study have laid a solid scientific foundation for large-scale quality control of pCms and have important theoretical and applied values.