Nature Identification of Chinese Herbal Medicine Compounds Based on Molecular Descriptors.

BACKGROUND Nature of Chinese Herbal Medicines (CHMs) is a bridge between traditional Chinese medicine (TCM) and clinical application. Accurate nature identification of CHMs is essential for guiding the clinical application of CHMs. OBJECTIVE To develop a new method for nature identification of CHMs according to compounds in CHMs. METHODS The nature of a CHM is a comprehensive manifestation of the properties of various compounds in this CHM. In this study, 2012 CHM compounds were extracted to construct a compound dataset. Molecular descriptors were utilized to build an identification model for classification of cold-hot-neutral nature of CHM compounds. RESULTS The predicting accuracy and confusion matrix were validated by the assembled dataset. The best model produced the accuracies of 96.5%±0.5% and 86.5%±1.5% on training set and test set, respectively. Furthermore, the identification model is robust in predicting hot-cold-neutral nature of the CHM compounds. CONCLUSIONS This work shows how a classification model for medical nature identification can be developed. The derived model can be utilized for the application of CHMs. HIGHLIGHTS To construct a nature identification model for analysis of hot-cold-neutral nature of CHMs according to the compounds in CHMs.