NormAE: Deep Adversarial Learning Model to Remove Batch Effects in Liquid Chromatography Mass Spectrometry-Based Metabolomics Data

Development and applications of metabolic models in plant multi-omics research

Plant growth and development are characterized by systematic and continuous processes, each involving intricate metabolic coordination mechanisms. Mathematical models are essential tools for investigating plant growth and development, metabolic regulation networks, and growth patterns across different stages. These models offer insights into secondary metabolism patterns in plants and the roles of metabolites. The proliferation of data related to plant genomics, transcriptomics, proteomics, and metabolomics in the last decade has underscored the growing importance of mathematical modeling in this field. This review aims to elucidate the principles and types of metabolic models employed in studying plant secondary metabolism, their strengths, and limitations. Furthermore, the application of mathematical models in various plant systems biology subfields will be discussed. Lastly, the review will outline how mathematical models can be harnessed to address research questions in this context.

Statistical analysis of feature-based molecular networking results from non-targeted metabolomics data

Feature-based molecular networking (FBMN) is a popular analysis approach for liquid chromatography-tandem mass spectrometry-based non-targeted metabolomics data. While processing liquid chromatography-tandem mass spectrometry data through FBMN is fairly streamlined, downstream data handling and statistical interrogation are often a key bottleneck. Especially users new to statistical analysis struggle to effectively handle and analyze complex data matrices. Here we provide a comprehensive guide for the statistical analysis of FBMN results, focusing on the downstream analysis of the FBMN output table. We explain the data structure and principles of data cleanup and normalization, as well as uni- and multivariate statistical analysis of FBMN results. We provide explanations and code in two scripting languages (R and Python) as well as the QIIME2 framework for all protocol steps, from data clean-up to statistical analysis. All code is shared in the form of Jupyter Notebooks ( https://github.com/Functional-Metabolomics-Lab/FBMN-STATS ). Additionally, the protocol is accompanied by a web application with a graphical user interface ( https://fbmn-statsguide.gnps2.org/ ) to lower the barrier of entry for new users and for educational purposes. Finally, we also show users how to integrate their statistical results into the molecular network using the Cytoscape visualization tool. Throughout the protocol, we use a previously published environmental metabolomics dataset for demonstration purposes. Together, the protocol, code and web application provide a complete guide and toolbox for FBMN data integration, cleanup and advanced statistical analysis, enabling new users to uncover molecular insights from their non-targeted metabolomics data. Our protocol is tailored for the seamless analysis of FBMN results from Global Natural Products Social Molecular Networking and can be easily adapted to other mass spectrometry feature detection, annotation and networking tools.

Machine learning for the advancement of genome-scale metabolic modeling

Constraint-based modeling (CBM) has evolved as the core systems biology tool to map the interrelations between genotype, phenotype, and external environment. The recent advancement of high-throughput experimental approaches and multi-omics strategies has generated a plethora of new and precise information from wide-ranging biological domains. On the other hand, the continuously growing field of machine learning (ML) and its specialized branch of deep learning (DL) provide essential computational architectures for decoding complex and heterogeneous biological data. In recent years, both multi-omics and ML have assisted in the escalation of CBM. Condition-specific omics data, such as transcriptomics and proteomics, helped contextualize the model prediction while analyzing a particular phenotypic signature. At the same time, the advanced ML tools have eased the model reconstruction and analysis to increase the accuracy and prediction power. However, the development of these multi-disciplinary methodological frameworks mainly occurs independently, which limits the concatenation of biological knowledge from different domains. Hence, we have reviewed the potential of integrating multi-disciplinary tools and strategies from various fields, such as synthetic biology, CBM, omics, and ML, to explore the biochemical phenomenon beyond the conventional biological dogma. How the integrative knowledge of these intersected domains has improved bioengineering and biomedical applications has also been highlighted. We categorically explained the conventional genome-scale metabolic model (GEM) reconstruction tools and their improvement strategies through ML paradigms. Further, the crucial role of ML and DL in omics data restructuring for GEM development has also been briefly discussed. Finally, the case-study-based assessment of the state-of-the-art method for improving biomedical and metabolic engineering strategies has been elaborated. Therefore, this review demonstrates how integrating experimental and in silico strategies can help map the ever-expanding knowledge of biological systems driven by condition-specific cellular information. This multiview approach will elevate the application of ML-based CBM in the biomedical and bioengineering fields for the betterment of society and the environment.

High-Resolution Mass Spectrometry for Human Exposomics: Expanding Chemical Space Coverage

In the modern "omics" era, measurement of the human exposome is a critical missing link between genetic drivers and disease outcomes. High-resolution mass spectrometry (HRMS), routinely used in proteomics and metabolomics, has emerged as a leading technology to broadly profile chemical exposure agents and related biomolecules for accurate mass measurement, high sensitivity, rapid data acquisition, and increased resolution of chemical space. Non-targeted approaches are increasingly accessible, supporting a shift from conventional hypothesis-driven, quantitation-centric targeted analyses toward data-driven, hypothesis-generating chemical exposome-wide profiling. However, HRMS-based exposomics encounters unique challenges. New analytical and computational infrastructures are needed to expand the analysis coverage through streamlined, scalable, and harmonized workflows and data pipelines that permit longitudinal chemical exposome tracking, retrospective validation, and multi-omics integration for meaningful health-oriented inferences. In this article, we survey the literature on state-of-the-art HRMS-based technologies, review current analytical workflows and informatic pipelines, and provide an up-to-date reference on exposomic approaches for chemists, toxicologists, epidemiologists, care providers, and stakeholders in health sciences and medicine. We propose efforts to benchmark fit-for-purpose platforms for expanding coverage of chemical space, including gas/liquid chromatography-HRMS (GC-HRMS and LC-HRMS), and discuss opportunities, challenges, and strategies to advance the burgeoning field of the exposome.

BERNN: Enhancing classification of Liquid Chromatography Mass Spectrometry data with batch effect removal neural networks

Liquid Chromatography Mass Spectrometry (LC-MS) is a powerful method for profiling complex biological samples. However, batch effects typically arise from differences in sample processing protocols, experimental conditions, and data acquisition techniques, significantly impacting the interpretability of results. Correcting batch effects is crucial for the reproducibility of omics research, but current methods are not optimal for the removal of batch effects without compressing the genuine biological variation under study. We propose a suite of Batch Effect Removal Neural Networks (BERNN) to remove batch effects in large LC-MS experiments, with the goal of maximizing sample classification performance between conditions. More importantly, these models must efficiently generalize in batches not seen during training. A comparison of batch effect correction methods across five diverse datasets demonstrated that BERNN models consistently showed the strongest sample classification performance. However, the model producing the greatest classification improvements did not always perform best in terms of batch effect removal. Finally, we show that the overcorrection of batch effects resulted in the loss of some essential biological variability. These findings highlight the importance of balancing batch effect removal while preserving valuable biological diversity in large-scale LC-MS experiments.

Data processing solutions to render metabolomics more quantitative: case studies in food and clinical metabolomics using Metabox 2.0

In classic semiquantitative metabolomics, metabolite intensities are affected by biological factors and other unwanted variations. A systematic evaluation of the data processing methods is crucial to identify adequate processing procedures for a given experimental setup. Current comparative studies are mostly focused on peak area data but not on absolute concentrations. In this study, we evaluated data processing methods to produce outputs that were most similar to the corresponding absolute quantified data. We examined the data distribution characteristics, fold difference patterns between 2 metabolites, and sample variance. We used 2 metabolomic datasets from a retail milk study and a lupus nephritis cohort as test cases. When studying the impact of data normalization, transformation, scaling, and combinations of these methods, we found that the cross-contribution compensating multiple standard normalization (ccmn) method, followed by square root data transformation, was most appropriate for a well-controlled study such as the milk study dataset. Regarding the lupus nephritis cohort study, only ccmn normalization could slightly improve the data quality of the noisy cohort. Since the assessment accounted for the resemblance between processed data and the corresponding absolute quantified data, our results denote a helpful guideline for processing metabolomic datasets within a similar context (food and clinical metabolomics). Finally, we introduce Metabox 2.0, which enables thorough analysis of metabolomic data, including data processing, biomarker analysis, integrative analysis, and data interpretation. It was successfully used to process and analyze the data in this study. An online web version is available at http://metsysbio.com/metabox.

SR-Unet: A Super-Resolution Algorithm for Ion Trap Mass Spectrometers Based on the Deep Neural Network

The mass spectrometer is an important tool for modern chemical analysis and detection. Especially, the emergence of miniature mass spectrometers has provided new tools for field analysis and detection. The resolution of a mass spectrometer reflects the ability of the instrument to discriminate between adjacent mass-to-charge ratio ions, and the higher the resolution, the better the discrimination of complex mixtures. Quadrupole ion traps are generally considered as a low-resolution mass spectrometry method, but they have gained wide attention and development in recent years because of their suitability for miniaturization and high qualitative capability. For an ion trap mass spectrometer, the mass sensitivity and resolution can be mutually constrained and need to be balanced by setting an appropriate scanning speed. In this study, a super-resolution U-net algorithm (SR-Unet) is proposed for ion trap mass spectrometry, which can estimate the possible ions from the overlapping ion peaks of low-resolution spectra and improve the equivalent resolution while ensuring sufficient sensitivity and analysis speed of the instrument. By determining the mass spectra of a linear ion trap mass spectrometer (LTQ XL) in Turbo and Normal scan modes, the same unit mass resolution as that at a scan speed of 16,667 Da/s was successfully obtained at 125,000 Da/s. Also, the experiments demonstrated that the algorithm is capable of the mass-to-charge ratio and instrument migration. SR-Unet can be migrated and applied to a miniature mass spectrometer for cruise detection of volatile organic compounds (VOCs), and the identification of VOC species in Photochemical Assessment Monitoring Stations (PAMS) was improved from 31 to 50 species with the same monitoring and analysis speed requirement. Further, super-unit mass resolution peptide detection was achieved on a miniature mass spectrometer with the help of the SR-Unet algorithm, which reduced the full width at half-maxima (FWHM) of bradykinin divalent ions (m/z 531) from 0.35 to 0.15 Da at a scan speed of 375 Da/s and improved the equivalent resolution to 3540. The proposed method provides a new idea to enhance the field mixture detection capability of miniature ion trap mass spectrometers.

Enhancing Classification of liquid chromatography mass spectrometry data with Batch Effect Removal Neural Networks (BERNN)

Liquid Chromatography Mass Spectrometry (LC-MS) is a powerful method for profiling complex biological samples. However, batch effects typically arise from differences in sample processing protocols, experimental conditions and data acquisition techniques, significantlyimpacting the interpretability of results. Correcting batch effects is crucial for the reproducibility of proteomics research, but current methods are not optimal for removal of batch effects without compressing the genuine biological variation under study. We propose a suite of Batch Effect Removal Neural Networks (BERNN) to remove batch effects in large LC-MS experiments, with the goal of maximizing sample classification performance between conditions. More importantly, these models must efficiently generalize in batches not seen during training. Comparison of batch effect correction methods across three diverse datasets demonstrated that BERNN models consistently showed the strongest sample classification performance. However, the model producing the greatest classification improvements did not always perform best in terms of batch effect removal. Finally, we show that overcorrection of batch effects resulted in the loss of some essential biological variability. These findings highlight the importance of balancing batch effect removal while preserving valuable biological diversity in large-scale LC-MS experiments.

Instrumental Drift in Untargeted Metabolomics: Optimizing Data Quality with Intrastudy QC Samples

Untargeted metabolomics is an important tool in studying health and disease and is employed in fields such as biomarker discovery and drug development, as well as precision medicine. Although significant technical advances were made in the field of mass-spectrometry driven metabolomics, instrumental drifts, such as fluctuations in retention time and signal intensity, remain a challenge, particularly in large untargeted metabolomics studies. Therefore, it is crucial to consider these variations during data processing to ensure high-quality data. Here, we will provide recommendations for an optimal data processing workflow using intrastudy quality control (QC) samples that identifies errors resulting from instrumental drifts, such as shifts in retention time and metabolite intensities. Furthermore, we provide an in-depth comparison of the performance of three popular batch-effect correction methods of different complexity. By using different evaluation metrics based on QC samples and a machine learning approach based on biological samples, the performance of the batch-effect correction methods were evaluated. Here, the method TIGER demonstrated the overall best performance by reducing the relative standard deviation of the QCs and dispersion-ratio the most, as well as demonstrating the highest area under the receiver operating characteristic with three different probabilistic classifiers (Logistic regression, Random Forest, and Support Vector Machine). In summary, our recommendations will help to generate high-quality data that are suitable for further downstream processing, leading to more accurate and meaningful insights into the underlying biological processes.

Circulating metabolites as potential biomarkers for the early detection and prognosis surveillance of gastrointestinal cancers

BACKGROUND AND AIMS

Two of the most lethal gastrointestinal (GI) cancers, gastric cancer (GC) and colon cancer (CC), are ranked in the top five cancers that cause deaths worldwide. Most GI cancer deaths can be reduced by earlier detection and more appropriate medical treatment. Unlike the current "gold standard" techniques, non-invasive and highly sensitive screening tests are required for GI cancer diagnosis. Here, we explored the potential of metabolomics for GI cancer detection and the classification of tissue-of-origin, and even the prognosis management.

METHODS

Plasma samples from 37 gastric cancer (GC), 17 colon cancer (CC), and 27 non-cancer (NC) patients were prepared for metabolomics and lipidomics analysis by three MS-based platforms. Univariate, multivariate, and clustering analyses were used for selecting significant metabolic features. ROC curve analysis was based on a series of different binary classifications as well as the true-positive rate (sensitivity) and the false-positive rate (1-specificity).

RESULTS

GI cancers exhibited obvious metabolic perturbation compared with benign diseases. The differentiated metabolites of gastric cancer (GC) and colon cancer (CC) were targeted to same pathways but with different degrees of cellular metabolism reprogramming. The cancer-specific metabolites distinguished the malignant and benign, and classified the cancer types. We also applied this test to before- and after-surgery samples, wherein surgical resection significantly altered the blood-metabolic patterns. There were 15 metabolites significantly altered in GC and CC patients who underwent surgical treatment, and partly returned to normal conditions.

CONCLUSION

Blood-based metabolomics analysis is an efficient strategy for GI cancer screening, especially for malignant and benign diagnoses. The cancer-specific metabolic patterns process the potential for classifying tissue-of-origin in multi-cancer screening. Besides, the circulating metabolites for prognosis management of GI cancer is a promising area of research.

Batch correction methods for nontarget chemical analysis data: application to a municipal wastewater collection system

Nontarget chemical analysis using high-resolution mass spectrometry has increasingly been used to discern spatial patterns and temporal trends in anthropogenic chemical abundance in natural and engineered systems. A critical experimental design consideration in such applications, especially those monitoring complex matrices over long time periods, is a choice between analyzing samples in multiple batches as they are collected, or in one batch after all samples have been processed. While datasets acquired in multiple analytical batches can include the effects of instrumental variability over time, datasets acquired in a single batch risk compound degradation during sample storage. To assess the influence of batch effects on the analysis and interpretation of nontarget data, this study examined a set of 56 samples collected from a municipal wastewater system over 7 months. Each month's samples included 6 from sites within the collection system, one combined influent, and one treated effluent sample. Samples were analyzed using liquid chromatography high-resolution mass spectrometry in positive electrospray ionization mode in multiple batches as the samples were collected and in a single batch at the conclusion of the study. Data were aligned and normalized using internal standard scaling and ComBat, an empirical Bayes method developed for estimating and removing batch effects in microarrays. As judged by multiple lines of evidence, including comparing principal variance component analysis between single and multi-batch datasets and through patterns in principal components and hierarchical clustering analyses, ComBat appeared to significantly reduce the influence of batch effects. For this reason, we recommend the use of more, small batches with an appropriate batch correction step rather than acquisition in one large batch.

Development of Non-Targeted Mass Spectrometry Method for Distinguishing Spelt and Wheat

Food fraud, even when not in the news, is ubiquitous and demands the development of innovative strategies to combat it. A new non-targeted method (NTM) for distinguishing spelt and wheat is described, which aids in food fraud detection and authenticity testing. A highly resolved fingerprint in the form of spectra is obtained for several cultivars of spelt and wheat using liquid chromatography coupled high-resolution mass spectrometry (LC-HRMS). Convolutional neural network (CNN) models are built using a nested cross validation (NCV) approach by appropriately training them using a calibration set comprising duplicate measurements of eleven cultivars of wheat and spelt, each. The results reveal that the CNNs automatically learn patterns and representations to best discriminate tested samples into spelt or wheat. This is further investigated using an external validation set comprising artificially mixed spectra, samples for processed goods (spelt bread and flour), eleven untypical spelt, and six old wheat cultivars. These cultivars were not part of model building. We introduce a metric called the D score to quantitatively evaluate and compare the classification decisions. Our results demonstrate that NTMs based on NCV and CNNs trained using appropriately chosen spectral data can be reliable enough to be used on a wider range of cultivars and their mixes.

A Comprehensive Mass Spectrometry-Based Workflow for Clinical Metabolomics Cohort Studies

As a comprehensive analysis of all metabolites in a biological system, metabolomics is being widely applied in various clinical/health areas for disease prediction, diagnosis, and prognosis. However, challenges remain in dealing with the metabolomic complexity, massive data, metabolite identification, intra- and inter-individual variation, and reproducibility, which largely limit its widespread implementation. This study provided a comprehensive workflow for clinical metabolomics, including sample collection and preparation, mass spectrometry (MS) data acquisition, and data processing and analysis. Sample collection from multiple clinical sites was strictly carried out with standardized operation procedures (SOP). During data acquisition, three types of quality control (QC) samples were set for respective MS platforms (GC-MS, LC-MS polar, and LC-MS lipid) to assess the MS performance, facilitate metabolite identification, and eliminate contamination. Compounds annotation and identification were implemented with commercial software and in-house-developed PAppLine^TM and Ulib^MS library. The batch effects were removed using a deep learning model method (NormAE). Potential biomarkers identification was performed with tree-based modeling algorithms including random forest, AdaBoost, and XGBoost. The modeling performance was evaluated using the F1 score based on a 10-times repeated trial for each. Finally, a sub-cohort case study validated the reliability of the entire workflow.

reComBat: batch-effect removal in large-scale multi-source gene-expression data integration

Motivation

With the steadily increasing abundance of omics data produced all over the world under vastly different experimental conditions residing in public databases, a crucial step in many data-driven bioinformatics applications is that of data integration. The challenge of batch-effect removal for entire databases lies in the large number of batches and biological variation, which can result in design matrix singularity. This problem can currently not be solved satisfactorily by any common batch-correction algorithm.

Results

We present reComBat, a regularized version of the empirical Bayes method to overcome this limitation and benchmark it against popular approaches for the harmonization of public gene-expression data (both microarray and bulkRNAsq) of the human opportunistic pathogen Pseudomonas aeruginosa. Batch-effects are successfully mitigated while biologically meaningful gene-expression variation is retained. reComBat fills the gap in batch-correction approaches applicable to large-scale, public omics databases and opens up new avenues for data-driven analysis of complex biological processes beyond the scope of a single study.

Availability and implementation

The code is available at https://github.com/BorgwardtLab/reComBat, all data and evaluation code can be found at https://github.com/BorgwardtLab/batchCorrectionPublicData.

Supplementary information

Supplementary data are available at Bioinformatics Advances online.

Recent advances in proteomics and metabolomics in plants

Over the past decade, systems biology and plant-omics have increasingly become the main stream in plant biology research. New developments in mass spectrometry and bioinformatics tools, and methodological schema to integrate multi-omics data have leveraged recent advances in proteomics and metabolomics. These progresses are driving a rapid evolution in the field of plant research, greatly facilitating our understanding of the mechanistic aspects of plant metabolisms and the interactions of plants with their external environment. Here, we review the recent progresses in MS-based proteomics and metabolomics tools and workflows with a special focus on their applications to plant biology research using several case studies related to mechanistic understanding of stress response, gene/protein function characterization, metabolic and signaling pathways exploration, and natural product discovery. We also present a projection concerning future perspectives in MS-based proteomics and metabolomics development including their applications to and challenges for system biology. This review is intended to provide readers with an overview of how advanced MS technology, and integrated application of proteomics and metabolomics can be used to advance plant system biology research.

Joint deep learning for batch effect removal and classification toward MALDI MS based metabolomics

Background

Metabolomics is a primary omics topic, which occupies an important position in both clinical applications and basic researches for metabolic signatures and biomarkers. Unfortunately, the relevant studies are challenged by the batch effect caused by many external factors. In last decade, the technique of deep learning has become a dominant tool in data science, such that one may train a diagnosis network from a known batch and then generalize it to a new batch. However, the batch effect inevitably hinders such efforts, as the two batches under consideration can be highly mismatched.

Results

We propose an end-to-end deep learning framework, for joint batch effect removal and then classification upon metabolomics data. We firstly validate the proposed deep learning framework on a public CyTOF dataset as a simulated experiment. We also visually compare the t-SNE distribution and demonstrate that our method effectively removes the batch effects in latent space. Then, for a private MALDI MS dataset, we have achieved the highest diagnostic accuracy, with about 5.1 ~ 7.9% increase on average over state-of-the-art methods.

Conclusions

Both experiments conclude that our method performs significantly better in classification than conventional methods benefitting from the effective removal of batch effect.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12859-022-04758-z.

Deep Learning Framework for Integrating Multibatch Calibration, Classification, and Pathway Activities

The amount of available biological data has exploded since the emergence of high-throughput technologies, which is not only revolting the way we recognize molecules and diseases but also bringing novel analytical challenges to bioinformatics analysis. In recent years, deep learning has become a dominant technique in data science. However, classification accuracy is plagued with domain discrepancy. Notably, in the presence of multiple batches, domain discrepancy typically happens between individual batches. Most pairwise adaptation approaches may be suboptimal as they fail to eliminate external factors across multiple batches and take the classification task into account simultaneously. We propose a joint deep learning framework for integrating batch effect removal, classification, and downstream pathway activities upon biological data. To this end, we validate it on two MALDI MS-based metabolomics datasets. We have achieved the highest diagnostic accuracy (ACC), with a notable ∼10% improvement over other methods. Overall, these results indicate that our approach removes batch effect more effectively than state-of-the-art methods and yields more accurate classification as well as biomarkers for smart diagnosis.

Data harmonisation for information fusion in digital healthcare: A state-of-the-art systematic review, meta-analysis and future research directions

Removing the bias and variance of multicentre data has always been a challenge in large scale digital healthcare studies, which requires the ability to integrate clinical features extracted from data acquired by different scanners and protocols to improve stability and robustness. Previous studies have described various computational approaches to fuse single modality multicentre datasets. However, these surveys rarely focused on evaluation metrics and lacked a checklist for computational data harmonisation studies. In this systematic review, we summarise the computational data harmonisation approaches for multi-modality data in the digital healthcare field, including harmonisation strategies and evaluation metrics based on different theories. In addition, a comprehensive checklist that summarises common practices for data harmonisation studies is proposed to guide researchers to report their research findings more effectively. Last but not least, flowcharts presenting possible ways for methodology and metric selection are proposed and the limitations of different methods have been surveyed for future research.

Norm ISWSVR: A Data Integration and Normalization Approach for Large-Scale Metabolomics

Large-scale and long-period metabolomics study is more susceptible to various sources of systematic errors, resulting in nonreproducibility and poor data quality. A reliable and robust batch correction method removes unwanted systematic variations and improves the statistical power of metabolomics data, which undeniably becomes an important issue for the quality control of metabolomics. This study proposed a novel data normalization and integration method, Norm ISWSVR. It is a two-step approach via combining the best-performance internal standard correction with support vector regression normalization, comprehensively removing the systematic and random errors and matrix effects. This method was investigated in three untargeted lipidomics or metabolomics datasets, and the performance was further evaluated systematically in comparison with that of 11 other normalization methods. As a result, Norm ISWSVR decreased the data's median cross-validated relative standard deviation (cvRSD), increased the correlation between QCs, improved the classification accuracy of biomarkers, and was well-compatible with quantitative data. More importantly, Norm ISWSVR also allows a low frequency of QCs, which could significantly decrease the burden of a large-scale experiment. Correspondingly, Norm ISWSVR favorably improves the data quality of large-scale metabolomics data.

Evaluating and minimizing batch effects in metabolomics

Determining metabolomic differences among samples of different phenotypes is a critical component of metabolomics research. With the rapid advances in analytical tools such as ultrahigh-resolution chromatography and mass spectrometry, an increasing number of metabolites can now be profiled with high quantification accuracy. The increased detectability and accuracy raise the level of stringiness required to reduce or control any experimental artifacts that can interfere with the measurement of phenotype-related metabolome changes. One of the artifacts is the batch effect that can be caused by multiple sources. In this review, we discuss the origins of batch effects, approaches to detect interbatch variations, and methods to correct unwanted data variability due to batch effects. We recognize that minimizing batch effects is currently an active research area, yet a very challenging task from both experimental and data processing perspectives. Thus, we try to be critical in describing the performance of a reported method with the hope of stimulating further studies for improving existing methods or developing new methods.

Deep Learning-Assisted Peak Curation for Large-Scale LC-MS Metabolomics

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 improves peak curation and contributes to the robust and scalable analysis of large-scale experiments. We show how to integrate it into different liquid chromatography–mass spectrometry (LC-MS) analysis workflows, quantify its performance, and compare it to various other approaches. NeatMS software is available as open source on github under permissive MIT license and is also provided as easy-to-install PyPi and Bioconda packages.

Image Processing and Machine Learning for Automated Identification of Chemo-/Biomarkers in Chromatography-Mass Spectrometry

We present a method named NPFimg, which automatically identifies multivariate chemo-/biomarker features of analytes in chromatography-mass spectrometry (MS) data by combining image processing and machine learning. NPFimg processes a two-dimensional MS map (m/z vs retention time) to discriminate analytes and identify and visualize the marker features. Our approach allows us to comprehensively characterize the signals in MS data without the conventional peak picking process, which suffers from false peak detections. The feasibility of marker identification is successfully demonstrated in case studies of aroma odor and human breath on gas chromatography-mass spectrometry (GC-MS) even at the parts per billion level. Comparison with the widely used XCMS shows the excellent reliability of NPFimg, in that it has lower error rates of signal acquisition and marker identification. In addition, we show the potential applicability of NPFimg to the untargeted metabolomics of human breath. While this study shows the limited applications, NPFimg is potentially applicable to data processing in diverse metabolomics/chemometrics using GC-MS and liquid chromatography-MS. NPFimg is available as open source on GitHub (http://github.com/poomcj/NPFimg) under the MIT license.

Deep learning meets metabolomics: a methodological perspective

Deep learning (DL), an emerging area of investigation in the fields of machine learning and artificial intelligence, has markedly advanced over the past years. DL techniques are being applied to assist medical professionals and researchers in improving clinical diagnosis, disease prediction and drug discovery. It is expected that DL will help to provide actionable knowledge from a variety of 'big data', including metabolomics data. In this review, we discuss the applicability of DL to metabolomics, while presenting and discussing several examples from recent research. We emphasize the use of DL in tackling bottlenecks in metabolomics data acquisition, processing, metabolite identification, as well as in metabolic phenotyping and biomarker discovery. Finally, we discuss how DL is used in genome-scale metabolic modelling and in interpretation of metabolomics data. The DL-based approaches discussed here may assist computational biologists with the integration, prediction and drawing of statistical inference about biological outcomes, based on metabolomics data.

Addressing the batch effect issue for LC/MS metabolomics data in data preprocessing

With the growth of metabolomics research, more and more studies are conducted on large numbers of samples. Due to technical limitations of the Liquid Chromatography–Mass Spectrometry (LC/MS) platform, samples often need to be processed in multiple batches. Across different batches, we often observe differences in data characteristics. In this work, we specifically focus on data generated in multiple batches on the same LC/MS machinery. Traditional preprocessing methods treat all samples as a single group. Such practice can result in errors in the alignment of peaks, which cannot be corrected by post hoc application of batch effect correction methods. In this work, we developed a new approach that address the batch effect issue in the preprocessing stage, resulting in better peak detection, alignment and quantification. It can be combined with down-stream batch effect correction methods to further correct for between-batch intensity differences. The method is implemented in the existing workflow of the apLCMS platform. Analyzing data with multiple batches, both generated from standardized quality control (QC) plasma samples and from real biological studies, the new method resulted in feature tables with better consistency, as well as better down-stream analysis results. The method can be a useful addition to the tools available for large studies involving multiple batches. The method is available as part of the apLCMS package. Download link and instructions are at https://mypage.cuhk.edu.cn/academics/yutianwei/apLCMS/.

Deep Learning Driven GC-MS Library Search and Its Application for Metabolomics

Preliminary compound identification and peak annotation in gas chromatography-mass spectrometry is usually made using mass spectral databases. There are a few algorithms that enable performing a search of a spectrum in a large mass spectral library. In many cases, a library search procedure returns a wrong answer even if a correct compound is contained in a library. In this work, we present a deep learning driven approach to a library search in order to reduce the probability of such cases. Machine learning ranking (learning to rank) is a class of machine learning and deep learning algorithms that perform a comparison (ranking) of objects. This work introduces the usage of deep learning ranking for small molecules identification using low-resolution electron ionization mass spectrometry. Instead of simple similarity measures for two spectra, such as the dot product or the Euclidean distance between vectors that represent spectra, a deep convolutional neural network is used. The deep learning ranking model outperforms other approaches and enables reducing a fraction of wrong answers (at rank-1) by 9-23% depending on the used data set. Spectra from the Golm Metabolome Database, Human Metabolome Database, and FiehnLib were used for testing the model.