Deep Learning-Assisted Peak Curation for Large-Scale LC-MS Metabolomics.
Ontology highlight
ABSTRACT:
SUBMITTER: Gloaguen Y
PROVIDER: S-EPMC8969107 | biostudies-literature | 2022 Mar
REPOSITORIES: biostudies-literature
Publications
Deep Learning-Assisted Peak Curation for Large-Scale LC-MS Metabolomics.
Gloaguen Yoann Y Kirwan Jennifer A JA Beule Dieter D
Analytical chemistry 20220315 12
Available automated methods for peak detection in untargeted metabolomics suffer from poor precision. We present NeatMS, which uses machine learning based on a convoluted neural network to reduce the number and fraction of false peaks. NeatMS comes with a pre-trained model representing expert knowledge in the differentiation of true chemical signal from noise. Furthermore, it provides all necessary functions to easily train new models or improve existing ones by transfer learning. Thus, the tool ...[more]
Similar Datasets
Project description:Untargeted metabolomics can detect more than 10 000 peaks in a single LC-MS run. The correspondence between these peaks and metabolites, however, remains unclear. Here, we introduce a Peak Annotation and Verification Engine (PAVE) for annotating untargeted microbial metabolomics data. The workflow involves growing cells in 13C and 15N isotope-labeled media to identify peaks from biological compounds and their carbon and nitrogen atom counts. Improved deisotoping and deadducting are enabled by algorithms that integrate positive mode, negative mode, and labeling data. To distinguish metabolites and their fragments, PAVE experimentally measures the response of each peak to weak in-source collision induced dissociation, which increases the peak intensity for fragments while decreasing it for their parent ions. The molecular formulas of the putative metabolites are then assigned based on database searching using both m/ z and C/N atom counts. Application of this procedure to Saccharomyces cerevisiae and Escherichia coli revealed that more than 80% of peaks do not label, i.e., are environmental contaminants. More than 70% of the biological peaks are isotopic variants, adducts, fragments, or mass spectrometry artifacts yielding ∼2000 apparent metabolites across the two organisms. About 650 match to a known metabolite formula based on m/ z and C/N atom counts, with 220 assigned structures based on MS/MS and/or retention time to match to authenticated standards. Thus, PAVE enables systematic annotation of LC-MS metabolomics data with only ∼4% of peaks annotated as apparent metabolites.
| S-EPMC6501219 | biostudies-literature
Project description:PurposeThe curation of images using human resources is time intensive but an essential step for developing artificial intelligence (AI) algorithms. Our goal was to develop and implement an AI algorithm for image curation in a high-volume setting. We also explored AI tools that will assist in deploying a tiered approach, in which the AI model labels images and flags potential mislabels for human review.DesignImplementation of an AI algorithm.ParticipantsSeven-field stereoscopic images from multiple clinical trials.MethodsThe 7-field stereoscopic image protocol includes 7 pairs of images from various parts of the central retina along with images of the anterior part of the eye. All images were labeled for field number by reading center graders. The model output included classification of the retinal images into 8 field numbers. Probability scores (0-1) were generated to identify misclassified images, with 1 indicating a high probability of a correct label.Main outcome measuresAgreement of AI prediction with grader classification of field number and the use of probability scores to identify mislabeled images.ResultsThe AI model was trained and validated on 17 529 images and tested on 3004 images. The pooled agreement of field numbers between grader classification and the AI model was 88.3% (kappa, 0.87). The pooled mean probability score was 0.97 (standard deviation [SD], 0.08) for images for which the graders agreed with the AI-generated labels and 0.77 (SD, 0.19) for images for which the graders disagreed with the AI-generated labels (P < 0.0001). Using receiver operating characteristic curves, a probability score of 0.99 was identified as a cutoff for distinguishing mislabeled images. A tiered workflow using a probability score of < 0.99 as a cutoff would include 27.6% of the 3004 images for human review and reduce the error rate from 11.7% to 1.5%.ConclusionsThe implementation of AI algorithms requires measures in addition to model validation. Tools to flag potential errors in the labels generated by AI models will reduce inaccuracies, increase trust in the system, and provide data for continuous model development.
| S-EPMC9754974 | biostudies-literature
Project description:Untargeted metabolomics is the study of all detectable small molecules, and in geroscience, metabolomics has shown great potential to describe the biological age-a complex trait impacted by many factors. Unfortunately, the sample sizes are often insufficient to achieve sufficient power and minimize potential biases caused by, for example, demographic factors. In this study, we present the analysis of biological age in ~10,000 toxicologic routine blood measurements. The untargeted screening samples obtained from ultra-high pressure liquid chromatography-quadruple time of flight mass spectrometry (UHPLC- QTOF) cover + 300 batches and + 30 months, lack pooled quality controls, lack controlled sample collection, and has previously only been used in small-scale studies. To overcome experimental effects, we developed and tested a custom neural network model and compared it with existing prediction methods. Overall, the neural network was able to predict the chronological age with an rmse of 5.88 years (r2 = 0.63) improving upon the 6.15 years achieved by existing normalization methods. We used the feature importance algorithm, Shapley Additive exPlanations (SHAP), to identify compounds related to the biological age. Most importantly, the model returned known aging markers such as kynurenine, indole-3-aldehyde, and acylcarnitines along with a potential novel aging marker, cyclo (leu-pro). Our results validate the association of tryptophan and acylcarnitine metabolism to aging in a highly uncontrolled large-s cale sample. Also, we have shown that by using robust computational methods it is possible to deploy large LC-MS datasets for metabolomics studies to reduce the risk of bias and empower aging studies.
| S-EPMC10186604 | biostudies-literature
Project description:MotivationPeak detection is a key step in the preprocessing of untargeted metabolomics data generated from high-resolution liquid chromatography-mass spectrometry (LC/MS). The common practice is to use filters with predetermined parameters to select peaks in the LC/MS profile. This rigid approach can cause suboptimal performance when the choice of peak model and parameters do not suit the data characteristics.ResultsHere we present a method that learns directly from various data features of the extracted ion chromatograms (EICs) to differentiate between true peak regions from noise regions in the LC/MS profile. It utilizes the knowledge of known metabolites, as well as robust machine learning approaches. Unlike currently available methods, this new approach does not assume a parametric peak shape model and allows maximum flexibility. We demonstrate the superiority of the new approach using real data. Because matching to known metabolites entails uncertainties and cannot be considered a gold standard, we also developed a probabilistic receiver-operating characteristic (pROC) approach that can incorporate uncertainties.Availability and implementationThe new peak detection approach is implemented as part of the apLCMS package available at http://web1.sph.emory.edu/apLCMS/ CONTACT: tyu8@emory.eduSupplementary informationSupplementary data are available at Bioinformatics online.
| S-EPMC4184266 | biostudies-literature
Project description:IntroductionDespite the availability of several pre-processing software, poor peak integration remains a prevalent problem in untargeted metabolomics data generated using liquid chromatography high-resolution mass spectrometry (LC-MS). As a result, the output of these pre-processing software may retain incorrectly calculated metabolite abundances that can perpetuate in downstream analyses.ObjectivesTo address this problem, we propose a computational methodology that combines machine learning and peak quality metrics to filter out low quality peaks.MethodsSpecifically, we comprehensively and systematically compared the performance of 24 different classifiers generated by combining eight classification algorithms and three sets of peak quality metrics on the task of distinguishing reliably integrated peaks from poorly integrated ones. These classifiers were compared to using a residual standard deviation (RSD) cut-off in pooled quality-control (QC) samples, which aims to remove peaks with analytical error.ResultsThe best performing classifier was found to be a combination of the AdaBoost algorithm and a set of 11 peak quality metrics previously explored in untargeted metabolomics and proteomics studies. As a complementary approach, applying our framework to peaks retained after filtering by 30% RSD across pooled QC samples was able to further distinguish poorly integrated peaks that were not removed from filtering alone. An R implementation of these classifiers and the overall computational approach is available as the MetaClean package at https://CRAN.R-project.org/package=MetaClean .ConclusionOur work represents an important step forward in developing an automated tool for filtering out unreliable peak integrations in untargeted LC-MS metabolomics data.
| S-EPMC7895495 | biostudies-literature
Project description:Metabolomics, understood as the science that manages the study of compounds from the metabolism, is an essential tool for deciphering metabolic changes in disease. The experiments rely on the use of high-throughput analytical techniques such as liquid chromatography coupled to mass spectrometry (LC-ToF MS). This hyphenation has brought positive aspects such as higher sensitivity, specificity and the extension of the metabolome coverage in a single run. The analysis of a high number of samples in a single batch is currently not always feasible due to technical and practical issues (i.e., a drop of the MS signal) which result in the MS stopping during the experiment obtaining more than a single sample batch. In this situation, careful data treatment is required to enable an accurate joint analysis of multi-batch data sets. This paper summarizes the analytical strategies in large-scale metabolomic experiments; special attention has been given to QC preparation troubleshooting and data treatment. Moreover, labeled internal standards analysis and their aim in data treatment, and data normalization procedures (intra- and inter-batch) are described. These concepts are exemplified using a cohort of 165 patients from a study in asthma.
| S-EPMC6918290 | biostudies-literature
Project description:IntroductionMetabolomics analysis based on liquid chromatography-mass spectrometry (LC-MS) has been a prevalent method in the metabolic field. However, accurately quantifying all the metabolites in large metabolomics sample cohorts is challenging. The analysis efficiency is restricted by the abilities of software in many labs, and the lack of spectra for some metabolites also hinders metabolite identification.ObjectivesDevelop software that performs semi-targeted metabolomics analysis with an optimized workflow to improve quantification accuracy. The software also supports web-based technologies and increases laboratory analysis efficiency. A spectral curation function is provided to promote the prosperity of homemade MS/MS spectral libraries in the metabolomics community.MethodsMetaPro is developed based on an industrial-grade web framework and a computation-oriented MS data format to improve analysis efficiency. Algorithms from mainstream metabolomics software are integrated and optimized for more accurate quantification results. A semi-targeted analysis workflow is designed based on the concept of combining artificial judgment and algorithm inference.ResultsMetaPro supports semi-targeted analysis workflow and functions for fast QC inspection and self-made spectral library curation with easy-to-use interfaces. With curated authentic or high-quality spectra, it can improve identification accuracy using different peak identification strategies. It demonstrates practical value in analyzing large amounts of metabolomics samples.ConclusionWe offer MetaPro as a web-based application characterized by fast batch QC inspection and credible spectral curation towards high-throughput metabolomics data. It aims to resolve the analysis difficulty in semi-targeted metabolomics.
| S-EPMC10250499 | biostudies-literature
Project description:MotivationWhen metabolites are analyzed by electrospray ionization (ESI)-mass spectrometry, they are usually detected as multiple ion species due to the presence of isotopes, adducts and in-source fragments. The signals generated by these degenerate features (along with contaminants and other chemical noise) obscure meaningful patterns in MS data, complicating both compound identification and downstream statistical analysis. To address this problem, we developed Binner, a new tool for the discovery and elimination of many degenerate feature signals typically present in untargeted ESI-LC-MS metabolomics data.ResultsBinner generates feature annotations and provides tools to help users visualize informative feature relationships that can further elucidate the underlying structure of the data. To demonstrate the utility of Binner and to evaluate its performance, we analyzed data from reversed phase LC-MS and hydrophilic interaction chromatography (HILIC) platforms and demonstrated the accuracy of selected annotations using MS/MS. When we compared Binner annotations of 75 compounds previously identified in human plasma samples with annotations generated by three similar tools, we found that Binner achieves superior performance in the number and accuracy of annotations while simultaneously minimizing the number of incorrectly annotated principal ions. Data reduction and pattern exploration with Binner have allowed us to catalog a number of previously unrecognized complex adducts and neutral losses generated during the ionization of molecules in LC-MS. In summary, Binner allows users to explore patterns in their data and to efficiently and accurately eliminate a significant number of the degenerate features typically found in various LC-MS modalities.Availability and implementationBinner is written in Java and is freely available from http://binner.med.umich.edu.Supplementary informationSupplementary data are available at Bioinformatics online.
| S-EPMC7828469 | biostudies-literature
Project description:Combining liquid chromatography-mass spectrometry (LC-MS)-based metabolomics experiments that were collected over a long period of time remains problematic due to systematic variability between LC-MS measurements. Until now, most normalization methods for LC-MS data are model-driven, based on internal standards or intermediate quality control runs, where an external model is extrapolated to the dataset of interest. In the first part of this article, we evaluate several existing data-driven normalization approaches on LC-MS metabolomics experiments, which do not require the use of internal standards. According to variability measures, each normalization method performs relatively well, showing that the use of any normalization method will greatly improve data-analysis originating from multiple experimental runs. In the second part, we apply cyclic-Loess normalization to a Leishmania sample. This normalization method allows the removal of systematic variability between two measurement blocks over time and maintains the differential metabolites. In conclusion, normalization allows for pooling datasets from different measurement blocks over time and increases the statistical power of the analysis, hence paving the way to increase the scale of LC-MS metabolomics experiments. From our investigation, we recommend data-driven normalization methods over model-driven normalization methods, if only a few internal standards were used. Moreover, data-driven normalization methods are the best option to normalize datasets from untargeted LC-MS experiments.
| S-EPMC3760460 | biostudies-literature
Project description:Recent mass spectrometry (MS)-based techniques enable deep proteome coverage with relative quantitative analysis, resulting in increased identification of very weak signals accompanied by increased data size of liquid chromatography (LC)-MS/MS spectra. However, the identification of weak signals using an assignment strategy with poorer performance results in imperfect quantification with misidentification of peaks and ratio distortions. Manually annotating a large number of signals within a very large dataset is not a realistic approach. In this study, therefore, we utilized machine learning algorithms to successfully extract a higher number of peptide peaks with high accuracy and precision. Our strategy evaluated each peak identified using six different algorithms; peptide peaks identified by all six algorithms (i.e., unanimously selected) were subsequently assigned as true peaks, which resulted in a reduction in the false-positive rate. Hence, exact and highly quantitative peptide peaks were obtained, providing better performance than obtained applying the conventional criteria or using a single machine learning algorithm.
| S-EPMC8642397 | biostudies-literature