Advanced Mass-Spectra-Based Machine Learning for Predicting the Toxicity of Traditional Chinese Medicines.

Area of Science:

• Pharmacology
• Toxicology
• Computational Chemistry

Background:

• Traditional Chinese medicine (TCM) offers health benefits but poses toxicological risks due to complex chemical profiles.
• Conventional quantitative structure-activity relationship (QSAR) models struggle with the intricate nature of TCM compounds.
• Addressing TCM toxicity is crucial for maximizing therapeutic efficacy and patient safety.

Purpose of the Study:

• To develop a novel method for predicting TCM toxicity using advanced analytical descriptors.
• To identify specific toxic components within TCM formulations.
• To explore the molecular mechanisms underlying TCM-induced toxicity.

Main Methods:

• Correlating TCM toxicity with analytical descriptors derived from electron ionization mass spectra (EI-MS) data.
• Employing interpretable machine learning models for toxicity prediction and component identification.
• Utilizing molecular dynamics (MD) simulations to investigate interactions between toxic components and protein targets (e.g., cytochrome P450 3A4).

Main Results:

• An optimal classification model achieved a balanced accuracy exceeding 0.74 in predicting TCM toxicity.
• Specific toxic components, including 13-hexyloxacyclotridec-10-en-2-one and loliolide, were identified.
• MD simulations provided insights into the interactions of toxic compounds with key proteins like CYP3A4.

Conclusions:

• Analytical descriptor-based machine learning offers a robust approach to predicting TCM toxicity.
• This method enhances understanding of TCM toxicological profiles, enabling safer clinical applications.
• The findings support the use of this approach for predicting toxicity in complex real-world mixtures.