臺灣大學: 資訊網路與多媒體研究所曾宇鳳; 周承復韓羽晨Harn, Yeu-ChernYeu-ChernHarn2013-03-222018-07-052013-03-222018-07-052011http://ntur.lib.ntu.edu.tw//handle/246246/251172辨識混合物溶液(例如來自藥草植物和生物體液)中的化學成分是一件耗時的工作。並且，辨識的結果受限於現有的質譜圖圖譜資料庫。我們在這篇論文中提出一個新的方法，稱為結構獵人，可以有效率且快速的預測混合物中的成分。這個新方法收集了大規模的天然物 (natural products)和利用一些已知在混合物中成分的成分作為”種子”。對於任一實驗，若儀器為液相層析質譜儀 (LC-MS)，結構獵人接收混合物樣品的峰表 (peak table)和種子資訊如名字或是化學結構，並運用這些資訊去預測此一實驗樣品中的化學成分。除此之外，結構獵人包含了天然物骨架關係資料庫，此資料庫包含各種天然物骨架，依天然物骨架的不同群組以及骨架彼此間的關係。骨架彼此間的關係包含：一、子骨架：子骨架亦包含了母骨架。二、手足骨架：兄弟骨架彼此具有同樣的母骨架，包含相同數量的環在骨架中，因此，手足骨架彼此骨架大小相近。母骨架，子骨架及兄弟骨架合稱為骨架清單，此清單的所有骨架會用來做為後續演算法的根基。結構獵人運用計算方式，考慮骨架的支鏈權重，以骨架清單中所有的骨架為主結構，權重支鏈為輔，組織可能的化學結構。因此，可以找出最有可能的天然物化學結構以對應混合物圖譜的多種成分。結構獵人考慮了結構在自然界的出現頻率來產生新結構。我們經由不同的天然物混合物樣品來驗證結構獵人，並證實結構獵人可以在樣品中找出傳統方法無法找到的結構及快速有系統地回報具有精確度的結果。Identification of individual chemical constituents in a mixture solution especially the medicinal plants is a time-consuming task. In addition, the identification results are limited with available reference mass spectra. Here we present a novel approach to efficiently and accurately predict individual components in a mixture by a relational database of natural products and a set of known reference compounds as “seeds”. For a natural products structural elucidation experiment from LC-MS analysis, Structure Hunter takes the peak table of the mixture sample from LC-MS spectra and seeds information, such as chemical structure scaffolds of the potential components in the mixture as input. Aside from the user inputs, Structure Hunter includes a natural products scaffold relationship database containing scaffolds, the major chemical scaffold classifications and the parent and sibling relationship between each scaffold of the natural products. A children scaffold contains chemical structure of its parent. The sibling scaffolds are scaffolds with same numbers of ring, similar size and same parent. The parent and sibling scaffolds of the seed along with seeds itself form the list of scaffolds for further comparison. Structure Hunter computationally formulates possible chemical structures by extending the list of scaffolds with a weighted list of side chains from analyzing the collections of natural products. Therefore, the multiple components in a mixture spectrum will be proposed by matching to the most probably chemical structures of the natural products and its analogous. Compared to the previous developed method with heuristics rules or chemical structural search to generate structure, Structure Hunter can generate structures relatively to their frequency in nature. Structure Hunter was proven to identify novel compounds and retrieve high-precision results with multiple natural products mixture samples.7289621 bytesapplication/pdfen-US質譜圖圖譜化學結構譜峰辨識天然物Liquid Chromatography-Mass Spectrometrynatural products structures elucidationpeak identificationcheminformaticschemometricsthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/251172/1/ntu-100-R98944018-1.pdf