Robust Automated Mass Spectra Interpretation and Chemical Formula Calculation Using Mixed Integer Linear Programming
Robust Automated Mass Spectra Interpretation and Chemical Formula Calculation Using Mixed Integer Linear Programming
复制标题
DOI：
10.1021/ac402180c
复制
发表时间：
2013-10-15
影响因子：
7.4
通讯作者：
Northen, Trent R.
中科院分区：
文献类型：
--
作者：
Baran, Richard;Northen, Trent R.
非靶向代谢物分析使用液相色谱和质谱结合电喷雾电离是发现新的天然产物，代谢能力和生物标志物的有力工具。然而，从光谱特征中阐明未表征代谢物的身份仍然具有挑战性。代谢物鉴定工作流程中的一个关键步骤是分配冗余光谱特征（加合物、片段、多聚物）和计算基本化学式。专家使用解决部分问题的计算工具（例如，单个离子的化学式计算）对数据进行检查，以消除替代解决方案的歧义，并提供可靠的结果。然而，手动管理是乏味的，而且不容易扩展或标准化。本文描述了一种基于混合整数线性规划优化（RAMSI）的鲁棒自动质谱解释和化学式计算的自动化过程。相关离子之间的化学规律被表示为线性约束，光谱解释和化学式计算在一个优化步骤中完成。这种方法是无偏的，因为它不需要预定义的中性损失集，并且正极性和负极性光谱可以在单个优化中组合。用30个实验质谱对该方法进行了评估，发现该方法可以有效地识别质子化或去质子化分子（[M + H]（+）或[M - H](-)），同时对背景离子的存在具有鲁棒性。RAMSI为在非靶向代谢组学工作流程中解释离子的后续鉴定提供了急需的标准化工具。
Untargeted metabolite profiling using liquid chromatography and mass spectrometry coupled via electrospray ionization is a powerful tool for the discovery of novel natural products, metabolic capabilities, and biomarkers. However, the elucidation of the identities of uncharacterized metabolites from spectral features remains challenging. A critical step in the metabolite identification workflow is the assignment of redundant spectral features (adducts, fragments, multimers) and calculation of the underlying chemical formula. Inspection of the data by experts using computational tools solving partial problems (e.g., chemical formula calculation for individual ions) can be performed to disambiguate alternative solutions and provide reliable results. However, manual curation is tedious and not readily scalable or standardized. Here we describe an automated procedure for the robust automated mass spectra interpretation and chemical formula calculation using mixed integer linear programming optimization (RAMSI). Chemical rules among related ions are expressed as linear constraints and both the spectra interpretation and chemical formula calculation are performed in a single optimization step. This approach is unbiased in that it does not require predefined sets of neutral losses and positive and negative polarity spectra can be combined in a single optimization. The procedure was evaluated with 30 experimental mass spectra and was found to effectively identify the protonated or deprotonated molecule ([M + H](+) or [M - H](-)) while being robust to the presence of background ions. RAMSI provides a much-needed standardized tool for interpreting ions for subsequent identification in untargeted metabolomics workflows.