Symposium Sessions
Day 2, June 11（Thu.） 10:30-10:45 Room B (4F 411+412)
- 2B-S1-1030（2P-06）
Investigation and Validation of Multivariate Analysis Methods for Estimating Natural Bioactive Compounds via Bioactivity-Based Molecular Networking
(1Gifu Univ., 2Gifu Univ., 3Gifu Univ.)
oShinsuke Takeuchi1, Kosei Yamauchi1,2,3
The discovery of natural bioactive compounds leads to the development of pharmaceutical seeds and functional ingredients. However, identifying such functional ingredients is not straightforward and requires considerable time and effort. In recent years, Bioactivity-Based Molecular Networking (BBMN) has gained attention as an efficient method for screening active compounds from natural product extracts. This approach integrates LC/MS profiles and bioactivity data on an MS/MS spectral similarity molecular network. Conventional BBMN employs univariate analysis using Pearson correlations between compound peak intensities and activity values. However, in multi-component systems, univariate analysis is highly susceptible to false positives arising from noise or variations in co-eluting components, making accurate estimation of activity contribution rates difficult. Therefore, this study introduced a multivariate analysis approach to construct OPLS-R (Orthogonal Partial Least Squares Regression) and OPLS-DA (Discriminant Analysis) models, aiming for robust and accurate estimation of active compounds. By integrating multiple statistical parameters (VIP, predictive loading value p(corr)) into the molecular network, we attempted to infer the active compounds in actual plant samples.