An integrated strategy combining metabolomics and machine learning for the evaluation of bioactive markers that differentiate various bile.

Animal bile is an important component of natural medicine and is widely used in clinical treatment. However, it is easy to cause mixed applications during processing, resulting in uneven quality, which seriously affects and harms the interests and health of consumers. Bile acids are the major bioactive constituents of bile and contain a variety of isomeric constituents. Although the components are structurally similar, they exhibit different pharmacological activities. Identifying the characteristics of each animal bile is particularly important for processing and reuse. It is necessary to establish an accurate analysis method to distinguish different types of animal bile. We evaluated the biological activity of key feature markers from various animal bile samples. In this study, a strategy combining metabolomics and machine learning was used to compare the bile of three different animals, and four key markers were screened. Quantitative analysis of the key markers showed that the levels of Glycochenodeoxycholic acid (GCDCA) and Taurodeoxycholic acid (TDCA) were highest in pig bile; Glycocholic acid (GCA) and Cholic acid (CA) were the most abundant in bovine and sheep bile, respectively. In addition, four key feature markers significantly inhibited the production of NO in LPS-stimulated RAW264.7 macrophage cells. These findings will contribute to the targeted development of bile in various animals and provide a basis for its rational application.

Comprehensive evaluation integrating omics strategy and machine learning algorithms for consistency of calculus bovis from different sources.

In the clinical application of Traditional Chinese Medicine (TCM) substitutes, the consistency evaluation of TCM substitutes from different sources is recognized as the main bottleneck. As the most widely used analytical method in TCM consistency evaluation, fingerprint similarity evaluation suffers from insufficient method sensitivity and poor conformity with the actual characteristics of TCM, which is difficult to adapt to the analytical needs of complex substance systems of TCM. This work aims to develop an effective and more accurate method for consistency evaluation using omics strategy and machine learning algorithms. The natural calculus bovis (NCB) were graded into three groups according to the similarity to in vitro cultured bovis (IVCB), and chemical markers between samples of each grade were screened out. Support vector machine (SVM) models with different kernels were then constructed by using the chemical markers as feature variables. The results showed that the classification accuracy of the SVM classifier of NCB and the consistency evaluation SVM model classifier was 95.74% and 100.0%, respectively. The approach demonstrated in the study presented a good analytical performance with higher sensitivity, accuracy for consistency evaluation of TCM.