Presentation Information
[P04-510]Comprehensive structural annotation of unidentified hydrophilic metabolites based on LC/HRMS/MS and in silico epimetabolite database (IEMDB)
○Taihei Torigoe1, Masatomo Takahashi1, Kohta Nakatani2, Yuki Soma3, Kosuke Hata1, Takeshi Bamba4, Yoshihiro Izumi1 (1. Mass Spectrometry Center, Graduate School of Science, The University of Osaka, Osaka, Japan (Japan), 2. Graduate School of Medical and Dental Sciences, Niigata University, Niigata, Japan (Japan), 3. National Institute of Advanced Industrial Science and Technology, Tsukuba, Japan (Japan), 4. Graduate School of Medicine, The University of Osaka, Osaka, Japan (Japan))
Keywords:
Epimetabolite,Structural annotation,E. coli
[Purpose]
Epimetabolites are defined as analogues of known metabolites with different substructures. Most epimetabolites remain unidentified due to the lack of a generally applicable method for their comprehensive annotation. Here, we propose an advanced methodology for the comprehensive structural elucidation of unidentified hydrophilic metabolites based on a combination of stable isotope labeling, unified-HILIC/AEX/HRMS/MS analysis, data mining techniques, and metabolite annotation using in silico epimetabolite database (IEMDB).
[Method]
We recently developed a unified hydrophilic-interaction/anion-exchange liquid chromatography tandem mass spectrometry (unified-HILIC/AEX/MS) technique that can comprehensively and simultaneously separate and detect a wide range of hydrophilic metabolites. The structural annotation accuracy of unidentified peaks obtained in liquid chromatography tandem mass spectrometry (LC/MS/MS) metabolomic analyses can be improved using retention time (RT) information as well as precursor and product ions. Therefore, we developed an RT prediction model for unified-HILIC/AEX, focusing on three key aspects: (1) acquiring and using a high-quality training dataset, (2) collecting of 12,420 molecular descriptors (explanatory variables), and (3) selecting molecular descriptors and constructing the model.For unknown molecular weight-related ion peaks that could not be identified or annotated using the in-house standard and public databases, a database of candidate structures predicted by in silico enzymatic reactions can be used. Up to two in silico metabolic reactions were performed using information from 56 known enzymatic reactions (e.g., oxidation and hydrolysis) and 2,244 hydrophilic compounds stored in the E. coli (ECMDB), yeast (YMDB), and human (HMDB) metabolite databases. As a result, 598,826 epimetabolite candidates were generated, and the IEMDB was constructed to register their chemical properties, including structural formula, SMILES, exact mass, etc. In addition, in silico MS/MS spectral information was integrated into the IEMDB for each epimetabolite candidate.
[Results]
The RT prediction model had an average absolute error of ±0.796 min relative to the test data. Additionally, 86% of the ΔRT (the difference between the measured and predicted values) were within ±1.5 min, and the R2 between the measured and predicted RTs was 0.92. We successfully annotated 216 metabolic features in untargeted metabolomics data of human plasma extract (SRM1950) using structural annotation via in silico MS/MS prediction and RT prediction. Notably, the number of candidate structures assigned by in silico MS/MS prediction was reduced by approximately 50% with the addition of RT prediction information.Furthermore, unified-HILIC/AEX/HRMS/MS analysis was performed on cell extracts obtained from E. coli cultured on either 12C6- or 13C6-glucose. The stable isotope labeling-based data mining technique (i.e., peak alignment, peak detection, and paired precursor ion filtering) accurately detected 2,046 hydrophilic metabolite candidates. A total of 106 of these metabolites were identified based on the in-house metabolite database. Finally, 44 new epimetabolites were successfully annotated by matching the retention time, precursor ion, and product ions of the epimetabolite candidates with those in IEMDB.
[Conclusion]
Our method is a potentially useful tool for elucidating the structure of epimetabolites.
Epimetabolites are defined as analogues of known metabolites with different substructures. Most epimetabolites remain unidentified due to the lack of a generally applicable method for their comprehensive annotation. Here, we propose an advanced methodology for the comprehensive structural elucidation of unidentified hydrophilic metabolites based on a combination of stable isotope labeling, unified-HILIC/AEX/HRMS/MS analysis, data mining techniques, and metabolite annotation using in silico epimetabolite database (IEMDB).
[Method]
We recently developed a unified hydrophilic-interaction/anion-exchange liquid chromatography tandem mass spectrometry (unified-HILIC/AEX/MS) technique that can comprehensively and simultaneously separate and detect a wide range of hydrophilic metabolites. The structural annotation accuracy of unidentified peaks obtained in liquid chromatography tandem mass spectrometry (LC/MS/MS) metabolomic analyses can be improved using retention time (RT) information as well as precursor and product ions. Therefore, we developed an RT prediction model for unified-HILIC/AEX, focusing on three key aspects: (1) acquiring and using a high-quality training dataset, (2) collecting of 12,420 molecular descriptors (explanatory variables), and (3) selecting molecular descriptors and constructing the model.For unknown molecular weight-related ion peaks that could not be identified or annotated using the in-house standard and public databases, a database of candidate structures predicted by in silico enzymatic reactions can be used. Up to two in silico metabolic reactions were performed using information from 56 known enzymatic reactions (e.g., oxidation and hydrolysis) and 2,244 hydrophilic compounds stored in the E. coli (ECMDB), yeast (YMDB), and human (HMDB) metabolite databases. As a result, 598,826 epimetabolite candidates were generated, and the IEMDB was constructed to register their chemical properties, including structural formula, SMILES, exact mass, etc. In addition, in silico MS/MS spectral information was integrated into the IEMDB for each epimetabolite candidate.
[Results]
The RT prediction model had an average absolute error of ±0.796 min relative to the test data. Additionally, 86% of the ΔRT (the difference between the measured and predicted values) were within ±1.5 min, and the R2 between the measured and predicted RTs was 0.92. We successfully annotated 216 metabolic features in untargeted metabolomics data of human plasma extract (SRM1950) using structural annotation via in silico MS/MS prediction and RT prediction. Notably, the number of candidate structures assigned by in silico MS/MS prediction was reduced by approximately 50% with the addition of RT prediction information.Furthermore, unified-HILIC/AEX/HRMS/MS analysis was performed on cell extracts obtained from E. coli cultured on either 12C6- or 13C6-glucose. The stable isotope labeling-based data mining technique (i.e., peak alignment, peak detection, and paired precursor ion filtering) accurately detected 2,046 hydrophilic metabolite candidates. A total of 106 of these metabolites were identified based on the in-house metabolite database. Finally, 44 new epimetabolites were successfully annotated by matching the retention time, precursor ion, and product ions of the epimetabolite candidates with those in IEMDB.
[Conclusion]
Our method is a potentially useful tool for elucidating the structure of epimetabolites.
Comment
To browse or post comments, you must log in.Log in