Structure Annotation of All Mass Spectra in Untargeted Metabolomics

Urine
metabolites are used in many clinical and biomedical studies
but usually only for a few classic compounds. Metabolomics detects
vastly more metabolic signals that may be used to precisely define
the health status of individuals. However, many compounds remain unidentified,
hampering biochemical conclusions. Here, we annotate all metabolites
detected by two untargeted metabolomic assays, hydrophilic interaction
chromatography (HILIC)-Q Exactive HF mass spectrometry and charged
surface hybrid (CSH)-Q Exactive HF mass spectrometry. Over 9,000 unique
metabolite signals were detected, of which 42% triggered MS/MS fragmentations
in data-dependent mode. On the highest Metabolomics Standards Initiative
(MSI) confidence level 1, we identified 175 compounds using authentic
standards with precursor mass, retention time, and MS/MS matching.
An additional 578 compounds were annotated by precursor accurate mass
and MS/MS matching alone, MSI level 2, including a novel library specifically
geared at acylcarnitines (CarniBlast). The rest of the metabolome
is usually left unannotated. To fill this gap, we used the in silico fragmentation tool CSI:FingerID and the new NIST
hybrid search to annotate all further compounds (MSI level 3). Testing
the top-ranked metabolites in CSI:Finger ID annotations yielded 40%
accuracy when applied to the MSI level 1 identified compounds. We
classified all MSI level 3 annotations by the NIST hybrid search using
the ClassyFire ontology into 21 superclasses that were further distinguished
into 184 chemical classes. ClassyFire annotations showed that the
previously unannotated urine metabolome consists of 28% derivatives
of organic acids, 16% heterocyclics, and 16% lipids as major classes.

尿液代谢物广泛应用于诸多临床与生物医学研究，但以往多仅针对数种经典化合物开展相关分析。代谢组学可检测到海量代谢信号，有望用于精准判定个体健康状态，但目前多数化合物仍未完成鉴定，这一现状阻碍了生化结论的推导。本研究针对两种非靶向代谢组学检测方法所捕获的全部代谢物进行注释，这两种方法分别为亲水相互作用色谱（hydrophilic interaction chromatography, HILIC）-Q Exactive HF 质谱与带电表面混合（charged surface hybrid, CSH）-Q Exactive HF 质谱。本研究共检测到超过9000个独特代谢物信号，其中42%在数据依赖性采集模式下触发了串联质谱（MS/MS）碎裂反应。基于最高等级的代谢组学标准倡议（Metabolomics Standards Initiative, MSI）置信水平1级，我们通过匹配前体离子质量、保留时间与串联质谱谱图并结合标准品对照，成功鉴定出175种化合物。另有578种化合物仅通过前体离子精确质量与串联质谱谱图匹配完成注释，对应MSI 2级，其中包含一款专门针对酰基肉碱类化合物的新型数据库（CarniBlast）。代谢组中剩余的绝大多数化合物通常未被注释。为填补这一空白，我们借助计算机模拟碎裂工具CSI:FingerID与新型NIST混合检索算法，对其余全部化合物完成注释（对应MSI 3级）。以MSI 1级已鉴定化合物作为验证集，对CSI:FingerID注释的高置信度代谢物进行测试，结果显示其准确率可达40%。我们依托ClassyFire本体论，通过NIST混合检索算法对所有MSI 3级注释结果进行分类，将其划分为21个超类，并进一步细分为184个化学类别。ClassyFire注释结果显示，此前未被注释的尿液代谢组主要由28%的有机酸衍生物、16%的杂环化合物以及16%的脂质构成。