基於質譜代謝體分析技術探討環境及生活習慣與人體健康之關聯性

Abstract

因為環境中的汙染物或特定生活習慣而造成的長期慢性暴露，往往帶來不良的身體健康影響。然而，人體、周遭環境及生活習慣間的關係是相當複雜且難以運用單一技術或儀器進行分析探討的。因此，我們將透過不同技術的整合探討三者間之關聯性。 生活環境會受到個人的行為及健康狀態影響。因此，分析室內灰塵的化學組成能讓我們更進一步了解人體健康與環境的關係。同時，為了瞭解人體與生活習慣之關聯，我們探討了嚼食檳榔及攝取保健食品之議題。我們應用了質譜儀分析多樣化的樣品並收集多樣且高質量的資訊。更利用液相層析質譜儀來獲得時間解析的化學資訊。除此之外，結合GNPS及機器學習等高通量數據處理方法，我們得以在短暫且有限的時間內分析龐大數據，並應用於其他型態的樣品上。 在本論文中，我們使用質譜儀並結合多樣分析數據方法，來探討以下兩大類的議題: (1) 利用紙片即時直接分析質譜儀顯著提升分析速度及定量之準確性；(2) 結合串聯式質譜儀及分析數據平台，如GNPS與監督式機器學習，了解人體、周遭環境及生活型態之關係。

There is a high chance for a human to have adverse health outcomes because of the long term chronic exposure under a polluted environment or having improper lifestyles. However, the interactions between the human, surrounding environment and the lifestyles are complicated and hard to fathom with only one single technique. Therefore, we demonstrated examples of clarifying these interactions by combining multiple techniques. The behaviors and conditions of individuals influence the environment. As a result, screening the composition of the chemicals of indoor dust allows us to get a deeper insight into the relationships between individuals and the environment. On the other hand, to understand the human health-habit interaction, we investigated the effect of chewing betel quid and intaking functional food. In our approach, we apply mass spectrometry (MS) to obtain diverse and high-quality information from various samples. Time-resolved chemical information was also obtained by tandem mass spectrometry technique. Furthermore, combining the MS data with high-throughput data processing strategies, such as GNPS and machine learning, we were able to analyze tremendous data within a little amount of time and to use the same framework in different samples. In this thesis, we used mass spectrometry combining analysis strategies to demonstrate that (1) the relationship between human, surrounding environment and the lifestyles can be understood by the combination of tandem mass spectrometry and data analysis platforms, including GNPS and supervised machine learning; (2) applying pDART significantly increase the analysis throughput while maintaining the quantitative accuracy compared to the traditional LC-MS approach.