統計代謝體學研究

Abstract

透過分析成千上百的代謝產物，代謝體數據分析提供很好的途徑來瞭解代謝機制，然而，代謝體數據的品質問題與複雜的批次效果至關重要，必須仰賴適當的統計分析來妥善解決。本研究開發了一套代謝體研究的整合分析工具，包含從最上游的數據前處理乃至下游的關聯性分析的整套完整分析流程。我們所開發的軟體Statistical Metabolomics Analysis – An R Tool（SMART）可以分析不同格式的代謝體數據，視覺化地呈現各種類型的數據，執行波峰調整和註釋，進行樣本與波峰的品質管制，偵測批次效果，進行關聯性分析，並完成波峰識別。使用SMART進行抗高血壓藥物的藥物代謝體研究分析，結果顯示代謝物神經介肽N（neuromedin N）對於血管緊張素轉換酶抑製劑有顯著的相關（在校正了未知潛在群體效果的ANCOVA分析中，得到p值 = 1.56 × 10−4；在校正了隱藏結構效果的ANCOVA分析中，得到p值 = 1.02 × 10−4）。此內生性神經肽（neuropeptide）與神經降壓素（neurotensin）和神經介肽U（neuromedin U）呈現高度相關，而這些蛋白質均參與調節血壓和平滑肌收縮的機制。SMART軟體程式、使用者指南和範例資料均可由http://www.stat.sinica.edu.tw/hsinchou/metabolomics/SMART.htm下載。

Metabolomics data provide unprecedented opportunities to decipher metabolic mechanisms by analyzing hundreds to thousands of metabolites. Data quality concerns and complex batch effects in metabolomics must be appropriately addressed through statistical analysis. This study developed an integrated analysis tool for metabolomics studies to streamline the complete analysis flow from initial data preprocessing to downstream association analysis. We developed Statistical Metabolomics Analysis – An R Tool (SMART), which can analyze input files with different formats, visually represent various types of data features, implement peak alignment and annotation, conduct quality control for replicate samples and peaks, explore batch effects, perform association analysis, and accomplish peak identification. A pharmacometabolomics study of antihypertensive medication was conducted and data were analyzed using SMART. Neuromedin N was identified as a metabolite significantly associated with angiotensin-converting-enzyme inhibitors in our metabolome-wide association analysis (p = 1.56 × 10−4 in an analysis of covariance (ANCOVA) with an adjustment for unknown latent groups and p = 1.02 × 10−4 in an ANCOVA with an adjustment for hidden substructures). This endogenous neuropeptide is highly related to neurotensin and neuromedin U, which are involved in blood pressure regulation and smooth muscle contraction. The SMART software, a user guide, and example data can be downloaded from http://www.stat.sinica.edu.tw/hsinchou/metabolomics/SMART.htm.