EVA: Evaluation of Metabolic Feature Fidelity Using a Deep Learning Model Trained With Over 25000 Extracted Ion Chromatograms

Guo, Jian; Shen, Sam; Xing, Shipei; Chen, Ying; Chen, Fanglin; Porter, Elizabeth; Yu, Huaxu; Huan, Tao

doi:10.1021/acs.analchem.1c01309

Cited by 26 publications

(27 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This somewhat worsens both the quality of the training and test data, which in turn reduces the prediction accuracy as can be seen in Figure S4 . However, the quality of peak integration affects all the machine learning models and therefore further developments in peak picking, grouping, and integration are highly needed to improve the quality of quantitative non-targeted analysis [ 42 ].…”

Section: Discussionmentioning

confidence: 99%

Machine Learning for Absolute Quantification of Unidentified Compounds in Non-Targeted LC/HRMS

Palm

Kruve

2022

Molecules

View full text Add to dashboard Cite

LC/ESI/HRMS is increasingly employed for monitoring chemical pollutants in water samples, with non-targeted analysis becoming more common. Unfortunately, due to the lack of analytical standards, non-targeted analysis is mostly qualitative. To remedy this, models have been developed to evaluate the response of compounds from their structure, which can then be used for quantification in non-targeted analysis. Still, these models rely on tentatively known structures while for most detected compounds, a list of structural candidates, or sometimes only exact mass and retention time are identified. In this study, a quantification approach was developed, where LC/ESI/HRMS descriptors are used for quantification of compounds even if the structure is unknown. The approach was developed based on 92 compounds analyzed in parallel in both positive and negative ESI mode with mobile phases at pH 2.7, 8.0, and 10.0. The developed approach was compared with two baseline approaches— one assuming equal response factors for all compounds and one using the response factor of the closest eluting standard. The former gave a mean prediction error of a factor of 29, while the latter gave a mean prediction error of a factor of 1300. In the machine learning-based quantification approach developed here, the corresponding prediction error was a factor of 10. Furthermore, the approach was validated by analyzing two blind samples containing 48 compounds spiked into tap water and ultrapure water. The obtained mean prediction error was lower than a factor of 6.0 for both samples. The errors were found to be comparable to approaches using structural information.

show abstract

Section: Discussionmentioning

confidence: 99%

Machine Learning for Absolute Quantification of Unidentified Compounds in Non-Targeted LC/HRMS

Palm

Kruve

2022

Molecules

View full text Add to dashboard Cite

show abstract

“…These datasets were downloaded from MetaboLights and Metabolomics Workbench. , After the universal parameters were generated in Paramounter, software-specific parameters were delivered to process these datasets in XCMS, MS-DIAL, and MZmine 2. The metabolic features in the resulting feature tables were then evaluated by EVA, a deep learning-based bioinformatic tool that can discriminate true positive metabolic features from false positives based on their chromatographic peak shapes . The total features and true positive features were both used to demonstrate the performance of Paramounter.…”

Section: Methodsmentioning

confidence: 99%

Paramounter: Direct Measurement of Universal Parameters To Process Metabolomics Data in a “White Box”

2022

Self Cite

View full text Add to dashboard Cite

Choosing appropriate data processing parameters is critical in processing liquid chromatography−mass spectrometry (LC−MS)-based untargeted metabolomics data. The conventional design of experiments (DOE) approach is time-consuming and provides no intuitive explanation why the selected parameters generate the best results. After studying commonly used metabolomics data processing software, this work summarized a set of universal parameters, including mass tolerance, peak height, peak width, and instrumental shift. These universal parameters are shared among different feature extraction programs and are critical to metabolic feature extraction. We then developed Paramounter, an R program that automatically measures these universal parameters from raw LC−MS-based metabolomics data prior to metabolic feature extraction. This is made possible through novel concepts of rank-based intensity sorting, zone of interest, and many others. Paramounter also translates universal parameters to software-specific parameters for data processing in different programs. Applying Paramounter is demonstrated to provide a threefold increase in the extracted metabolites compared to using default parameters in MS-DIAL-based feature extraction. Furthermore, the comparison between Paramounter, AutoTuner, and IPO showed that Paramounter generates 3.7-and 1.6-fold more true positive features than AutoTuner and IPO, respectively. Further validation of Paramounter on 11 datasets covering different sample types, data acquisition modes, and MS vendors proved that Paramounter is a convenient and robust program. Overall, the proposed universal parameters and the development of Paramounter address a critical need in metabolomics data processing, transforming metabolomics feature extraction from a "black box" to a "white box." Paramounter is freely available on GitHub (https://github.com/HuanLab/Paramounter).

show abstract

“…Representative EICs of metabolic features from both the metabolite and exposome standards generated by JPA-PP, JPA-MR, and JPA-TL are presented in Figure S5 . Feature fidelity was evaluated using EVA, a deep learning model trained with over 25,000 manually inspected EICs [ 30 ].…”

Section: Methodsmentioning

confidence: 99%

JPA: Joint Metabolic Feature Extraction Increases the Depth of Chemical Coverage for LC-MS-Based Metabolomics and Exposomics

et al. 2022

Self Cite

View full text Add to dashboard Cite

Extracting metabolic features from liquid chromatography-mass spectrometry (LC-MS) data has been a long-standing bioinformatic challenge in untargeted metabolomics. Conventional feature extraction algorithms fail to recognize features with low signal intensities, poor chromatographic peak shapes, or those that do not fit the parameter settings. This problem also poses a challenge for MS-based exposome studies, as low-abundant metabolic or exposomic features cannot be automatically recognized from raw data. To address this data processing challenge, we developed an R package, JPA (short for Joint Metabolomic Data Processing and Annotation), to comprehensively extract metabolic features from raw LC-MS data. JPA performs feature extraction by combining a conventional peak picking algorithm and strategies for (1) recognizing features with bad peak shapes but that have tandem mass spectra (MS2) and (2) picking up features from a user-defined targeted list. The performance of JPA in global metabolomics was demonstrated using serial diluted urine samples, in which JPA was able to rescue an average of 25% of metabolic features that were missed by the conventional peak picking algorithm due to dilution. More importantly, the chromatographic peak shapes, analytical accuracy, and precision of the rescued metabolic features were all evaluated. Furthermore, owing to its sensitive feature extraction, JPA was able to achieve a limit of detection (LOD) that was up to thousands of folds lower when automatically processing metabolomics data of a serial diluted metabolite standard mixture analyzed in HILIC(−) and RP(+) modes. Finally, the performance of JPA in exposome research was validated using a mixture of 250 drugs and 255 pesticides at environmentally relevant levels. JPA detected an average of 2.3-fold more exposure compounds than conventional peak picking only.

show abstract

EVA: Evaluation of Metabolic Feature Fidelity Using a Deep Learning Model Trained With Over 25000 Extracted Ion Chromatograms

Cited by 26 publications

References 22 publications

Machine Learning for Absolute Quantification of Unidentified Compounds in Non-Targeted LC/HRMS

Machine Learning for Absolute Quantification of Unidentified Compounds in Non-Targeted LC/HRMS

Paramounter: Direct Measurement of Universal Parameters To Process Metabolomics Data in a “White Box”

JPA: Joint Metabolic Feature Extraction Increases the Depth of Chemical Coverage for LC-MS-Based Metabolomics and Exposomics

Contact Info

Product

Resources

About