Page 175 - 2025中醫藥與天然藥物聯合學術研討會-中醫藥與天然藥物的挑戰Ｘ機遇與未來大會手冊

PC-57

Metabolic profiling of edible plant extracts with insights and limitations of

conventional bioinformatics

5,6
1,2
3,4
Justin Chang, Ho-Cheng Wu, Jacky Chung-Hao Wu, Henry Horng-Shing Lu, 5,6,7,8
Chia-Hung Yen* ,1,4,9

1 Graduate Institute of Natural Products, College of Pharmacy, Kaohsiung Medical
University, Kaohsiung 807378, Taiwan
2 Department of Biotechnology, College of Life Sciences, Indonesia International Institute of
Life-Sciences (i3L), Jakarta 13210, Indonesia
3 School of Pharmacy, College of Pharmacy, Kaohsiung Medical University, Kaohsiung
807378, Taiwan
4 Drug Development and Value Creation Research Center, Kaohsiung Medical University,
Kaohsiung 807378, Taiwan
5 Biomedical Artificial Intelligence Academy, Kaohsiung Medical University, Kaohsiung
807378, Taiwan
6 Institute of Statistics, National Yang Ming Chiao Tung University, Hsinchu 300093, Taiwan
7 Department of Medical Research, Kaohsiung Medical University Hospital, Kaohsiung
807378, Taiwan
8 Department of Statistics and Data Science, Cornell University, Ithaca, NY 14853, USA
9 Center for Cancer Research, Kaohsiung Medical University, Kaohsiung 807378, Taiwan
* E-mail: chyen@kmu.edu.tw

Abstract
Natural products remain an important source of new drugs, and metabolomics using MS/MS
analysis enables comprehensive profiling of complex natural extracts. Conventional
bioinformatics approaches such as PCA and heatmaps are widely used as standard tools for
visualization and exploratory analysis in metabolomics. They are valuable for revealing global
chemical patterns, yet their ability to resolve subtle features linked to specific bioactivities
remains uncertain. From the Natural Product Libraries and High-Throughput Screening Core
(NPS), Kaohsiung Medical University (KMU), 100 edible plant extracts with completed
MS/MS analysis and NRF2 inhibition testing were examined. Three active extracts and six non-
active extracts from the same families (1:2 ratio) were selected for comparative profiling.
MS/MS data were annotated using SIRIUS, and bioinformatics analyses including PCA and
heatmaps were performed based on compound identity and classification. The analyses revealed
that active and non-active extracts could not be clearly separated, regardless of the parameters
used. This indicates that while conventional bioinformatics approaches provide useful
metabolomic overviews, they may be insufficient to uncover activity-associated features. Our
findings highlight the need for more advanced computational strategies in future studies to
better link chemical signatures with NRF2 inhibition.

Keywords: Natural products; Metabolomics; MS/MS analysis; NRF2 inhibition;
Bioinformatics