代謝體是生命體細胞代謝過程中所產生的小分子集合，能反映生物體在特定生理或病理狀態下的體內生化反應。近年來，代謝體已廣泛應用於生物標記開發、藥物代謝機轉分析與疾病特徵分析等領域，顯示其在生物醫學研究領域中的重要地位。在眾多代謝體分析平台中，液相層析串聯質譜儀（Liquid Chromatography-Tandem Mass Spectrometry， LC-MS/MS）因其具高靈敏度、對多樣性代謝物之良好分析能力等特點，已成為非標靶代謝體學中最重要的分析工具之一。然而，在使用液相層析串聯質譜儀分析大規模跨批次樣本時，資料常受到批次效應（Batch effect）的影響，導致實驗結果的可比較性與準確性大幅降低，因此如何有效控制批次效應的影響在代謝體質譜儀資料分析中尤為重要。在現有的批次效應校正流程中，已經有許多批次效應校正演算法被開發出來，但是並不是每個演算法的校正效果都非常好。為了提升批次效應校正效果，本研究提出一套液相層析雜訊消除演算法CNEBat（Chromatographic Noise Elimination for Batch Effect Correction），以去除液相層析過程因批次差異所造成之系統性偏差，並作為批次效應校正前之處理步驟，以提升後續批次效應校正之表現。CNEBat在雜訊估計與校正中，使用單一樣本之mzML原始資料分別建立各樣本之雜訊校正回歸模型，而非以整批樣本共同建立單一校正模型。為了評估CNEBat之效能，本研究使用多組不同性質之代謝體資料進行分析，包括脂質代謝體品質控制資料集、惡性瘧原蟲細胞代謝體資料集、老鼠大腦脂質代謝體與老鼠大腦極性代謝體資料集。於主成分分析（Principal Components Analysis，PCA）結果中，經CNEBat處理並結合現有批次效應校正方法後，樣本於批次間之分群現象明顯降低；在單一特徵峰強度分析中，不同批次之內部標準品及已知代謝物之強度分布穩定性在特定演算法中能獲得提升；在品質控制樣本皮爾森積差相關係數分析中，不同批次樣本之間之相關性明顯提升。前述結果顯示，經CNEBat前處理並結合現有批次效應校正方法後，資料之批次效應校正效果獲得有效改善。 綜合上述結果，CNEBat 能有效降低液相層析過程所造成之系統性雜訊，並可作為批次效應校正前處理之方法，提升後續分析之穩定性與可靠性。;Metabolomics represents the ultimate outcome of biological gene expression， providing the most intuitive reflection of biochemical reactions occurring within an organism. Compared to upstream products that are influenced by epigenomics， this characteristic has led to the increasing significance of metabolomics. Among various metabolomics analysis platforms， liquid chromatography-tandem mass spectrometry (LC-MS/MS) plays a crucial role in untargeted metabolomics research. However， when analyzing large-scale cross-batch samples using LC-MS/MS， batch effects often pose a significant challenge. Effectively correcting batch effects to ensure the accuracy of experimental data has thus become a critical issue. To address this challenge， we have proposed a liquid chromatography noise elimination algorithm named CNEBat (Chromatographic Noise Elimination for Batch Effect Correction)， designed to eliminate batch-induced errors occurring during the liquid chromatography (LC) process. Implemented in Java， CNEBat supports cross-platform operation and provides a highly compatible， user-friendly solution for liquid chromatography noise elimination. It performs noise calculation and removal using mzML data from a single sample， thereby preventing error correction caused by extreme values and ensuring optimal correction for each sample. After correction with CNEBat， users can further integrate the method with other batch effect correction algorithms as needed to achieve enhanced correction performance. Moreover， CNEBat operates solely based on mzML files and a user-defined Internal Standard List， allowing users to store computation coefficients from a single dataset. If users subsequently process data with other quantitative software， they can apply the same set of computation coefficients for rapid calculations， thereby improving compatibility across different quantification platforms and significantly reducing computational time.
