A data preprocessing strategy for metabolomics to reduce the mask effect in data analysis

Lipopolysaccharide perception in Arabidopsis thaliana: Diverse LPS chemotypes from Burkholderia cepacia, Pseudomonas syringae and Xanthomonas campestris trigger differential defence-related perturbations in the metabolome

Lipopolysaccharides (LPSs) are microbe-associated molecular pattern molecules (MAMPs) from Gram-negative bacterial pathogens that potentially contain three different MAMPs (the O-polysaccharide chain, the oligosaccharide core and lipid A). LPSs was purified from Burkholderia cepacia, Pseudomonas syringae and Xanthomonas campestris and electrophoretically profiled. Outcomes of the interactions of the three different LPS chemotypes with Arabidopsis thaliana, as reflected in the induced defence metabolites, profiled at 12 h and 24 h post elicitation, were investigated. Plants were pressure-infiltrated with LPS solutions and methanol-based extractions at different time points were performed for untargeted metabolomics using ultra-high performance liquid chromatography coupled to quadrupole time-of-flight mass spectrometry. Multivariate data modelling and chemometric analysis were applied to generate interpretable biochemical information from the multidimensional data sets. The three LPSs triggered differential metabolome changes in the plants as apparent from chromatographically distinct MS chromatograms. Unsupervised and supervised multivariate data models exhibited time- and treatment-related variations, and revealed discriminating metabolite variables. Heat map models comparatively displayed the up-regulated pathways affecting the metabolomes and Venn diagrams indicated up-regulated and shared metabolites among the three LPS treatments. The altered metabolomes reflect the up-regulation of metabolites from not only the glucosinolate pathway, but also from the shikimate-phenylpropanoid-flavonoid -, terpenoid - and indolic/alkaloid pathways, as well as oxygenated fatty acids. Distinct phytochemical profiles, especially at the earlier time point, suggest differences in the perception of the three LPS chemotypes, associated with the molecular patterns within the tripartite lipoglycans.

An improved pseudotargeted GC-MS/MS-based metabolomics method and its application in radiation-induced hepatic injury in a rat model

The liver is the pivotal metabolic organ primarily responsible for metabolic activities, detoxification and regulation of carbohydrate, protein, amino acid, and lipid metabolism. However, very little is known about the complicated pathophysiologic mechanisms of liver injury result from ionizing radiation exposure. Therefore, a pseudotargeted metabolomics approach based on gas chromatography-tandem mass spectrometry with selected reaction monitoring (GC-MS-SRM) was developed to study metabolic alterations of liver tissues in radiation-induced hepatic injury. The pseudotargeted GC-MS-SRM method was validated with satisfactory analytical characteristics in terms of precision, linearity, sensitivity and recovery. Compared to the SIM-based approach, the SRM scanning method had mildly better precision, higher sensitivity, and wider linear ranges. A total of 37 differential metabolites associated with radiation-induced hepatic injury were identified using the GC-MS-SRM metabolomics method. Global metabolic clustering analysis showed that amino acids, carbohydrates, unsaturated fatty acids, organic acids, metabolites associated with pyrimidine metabolism, ubiquinone biosynthesis and oxidative phosphorylation appeared significantly declined after high dose irradiation exposure, whereas metabolites related to lysine catabolism, glycerolipid metabolism and glutathione metabolism presented the opposite behavior. These changes indicate energy deficiency, antioxidant defense damage, accumulation of ammonia and lipid oxidation of liver tissues in response to radiation exposure. It is shown that the developed pseudotargeted method based on GC-MS-SRM is a useful tool for metabolomics study.

Wood Metabolomic Responses of Wild and Cultivated Grapevine to Infection with Neofusicoccum parvum, a Trunk Disease Pathogen

Grapevine trunk diseases (GTDs), which are associated with complex of xylem-inhabiting fungi, represent one of the major threats to vineyard sustainability currently. Botryosphaeria dieback, one of the major GTDs, is associated with wood colonization by Botryosphaeriaceae fungi, especially Neofusicoccum parvum. We used GC-MS and HPLC-MS to compare the wood metabolomic responses of the susceptible Vitis vinifera subsp. vinifera (V. v. subsp. vinifera) and the tolerant Vitis vinifera subsp. sylvestris (V. v. subsp. sylvestris) after artificial inoculation with Neofusicoccum parvum (N. parvum). N. parvum inoculation triggered major changes in both primary and specialized metabolites in the wood. In both subspecies, infection resulted in a strong decrease in sugars (fructose, glucose, sucrose), whereas sugar alcohol content (mannitol and arabitol) was enhanced. Concerning amino acids, N. parvum early infection triggered a decrease in aspartic acid, serine, and asparagine, and a strong increase in alanine and β-alanine. A trend for more intense primary metabolism alteration was observed in V. v. subsp. sylvestris compared to V. v. subsp. vinifera. N. parvum infection also triggered major changes in stilbene and flavonoid compounds. The content in resveratrol and several resveratrol oligomers increased in the wood of both subspecies after infection. Interestingly, we found a higher induction of resveratrol oligomer (putative E-miyabenol C, vitisin C, hopeaphenol, ampelopsin C) contents after wood inoculation in V. v. subsp. sylvestris.

Application of metabolomics to the study of irritable bowel syndrome

The pathophysiology of irritable bowel syndrome and the detection of biomarkers of specific mechanisms and/or predictors of therapeutic response remain elusive. This roadblock reflects, in large part, the complexity and heterogeneity of the disorder. Recently, there has been growing evidence of a dietary and/or microbiome interaction with the host that may trigger symptoms in a subset of patients. While a number of techniques are available to examine these potential interactions, "omic" approaches such as metabolomics are becoming more widely used. Metabolomics measures hundreds and potentially thousands of known and unknown small molecule chemicals (metabolites) to provide a unique look into mechanisms that underlie symptom generation and potential predictors of therapeutic response. In this issue of the journal, Lee et al use nuclear magnetic resonance (NMR) to demonstrate the value of this approach to study IBS. This review examines the use of metabolomics to better understand IBS, focusing on what has been learned to date, practical and technical considerations, its potential for future research and how the study by Lee et al have contributed to these concepts.

The use of missing values in proteomic data-independent acquisition mass spectrometry to enable disease activity discrimination

MOTIVATION

Data-independent acquisition mass spectrometry allows for comprehensive peptide detection and relative quantification than standard data-dependent approaches. While less prone to missing values, these still exist. Current approaches for handling the so-called missingness have challenges. We hypothesized that non-random missingness is a useful biological measure and demonstrate the importance of analysing missingness for proteomic discovery within a longitudinal study of disease activity.

RESULTS

The magnitude of missingness did not correlate with mean peptide concentration. The magnitude of missingness for each protein strongly correlated between collection time points (baseline, 3 months, 6 months; R = 0.95-0.97, confidence interval = 0.94-0.97) indicating little time-dependent effect. This allowed for the identification of proteins with outlier levels of missingness that differentiate between the patient groups characterized by different patterns of disease activity. The association of these proteins with disease activity was confirmed by machine learning techniques. Our novel approach complements analyses on complete observations and other missing value strategies in biomarker prediction of disease activity.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

The ABRF Metabolomics Research Group 2016 Exploratory Study: Investigation of Data Analysis Methods for Untargeted Metabolomics

Lack of standardized applications of bioinformatics and statistical approaches for pre- and postprocessing of global metabolomic profiling data sets collected using high-resolution mass spectrometry platforms remains an inadequately addressed issue in the field. Several publications now recognize that data analysis outcome variability is caused by different data treatment approaches. Yet, there is a lack of interlaboratory reproducibility studies that have looked at the contribution of data analysis techniques toward variability/overlap of results. The goal of our study was to identify the contribution of data pre- and postprocessing methods on metabolomics analysis results. We performed urinary metabolomics from samples obtained from mice exposed to 5 Gray of external beam gamma rays and those exposed to sham irradiation (control group). The data files were made available to study participants for comparative analysis using commonly used bioinformatics and/or biostatistics approaches in their laboratories. The participants were asked to report back the top 50 metabolites/features contributing significantly to the group differences. Herein we describe the outcome of this study which suggests that data preprocessing is critical in defining the outcome of untargeted metabolomic studies.

The association of Chlamydia trachomatis and Mycoplasma genitalium infection with the vaginal metabolome

Chlamydia trachomatis (CT) and Mycoplasma genitalium (MG) are two highly prevalent bacterial sexually transmitted infections (STIs) with a significant rate of co-infection in some populations. Vaginal metabolites are influenced by resident vaginal microbiota, affect susceptibility to sexually transmitted infections (STIs), and may impact local inflammation and patient symptoms. Examining the vaginal metabolome in the context of CT mono (CT+) and CT/MG co-infection (CT+/MG+) may identify biomarkers for infection or provide new insights into disease etiology and pathogenesis. Yet, the vaginal metabolome in the setting of CT infection is understudied and the composition of the vaginal metabolome in CT/MG co-infected women is unknown. Therefore, in this analysis, we used an untargeted metabolomic approach combined with 16S rRNA gene amplicon sequencing to characterize the vaginal microbiota and metabolomes of CT+, CT+/MG+, and uninfected women. We found that CT+ and CT+/MG+ women had distinct vaginal metabolomic profiles as compared to uninfected women both before and after adjustment for the vaginal microbiota. This study provides important foundational data documenting differences in the vaginal metabolome between CT+, CT+/MG+ and uninfected women. These data may guide future mechanistic studies that seek to provide insight into the pathogenesis of CT and CT/MG infections.

Potential Impacts of Translocation of Neonicotinoid Insecticides to Cotton (Gossypium hirsutum (Malvales: Malvaceae)) Extrafloral Nectar on Parasitoids

Neonicotinoid seed treatments are frequently used in cotton (Gossypium hirsutum L. [Malvales: Malvaceae]) production to provide protection against early-season herbivory. However, there is little known about how these applications affect extrafloral nectar (EFN), an important food resource for arthropod natural enemies. Using enzyme-linked immunosorbent assays, we found that neonicotinoids were translocated to the EFN of clothianidin- and imidacloprid-treated, greenhouse-grown cotton plants at concentrations of 77.3 ± 17.3 and 122.6 ± 11.5 ppb, respectively. We did not find differences in the quantity of EFN produced by neonicotinoid-treated cotton plants compared to untreated controls, either constitutively or after mechanical damage. Metabolomic analysis of sugars and amino acids from treated and untreated plants did not detect differences in overall composition of EFN. In bioassays, female Cotesia marginiventris (Cresson) (Hymenoptera: Braconidae) parasitoid wasps that fed on EFN from untreated, clothianidin-treated, or imidacloprid-treated plants demonstrated no difference in mortality or parasitization success. We also conducted acute toxicity assays for C. marginiventris fed on honey spiked with clothianidin and imidacloprid and established LC50 values for male and female wasps. Although LC50 values were substantially higher than neonicotinoid concentrations detected in EFN, caution should be used when translating these results to the field where other stressors could alter the effects of neonicotinoids. Moreover, there are a wide range of possible sublethal impacts of neonicotinoids, none of which were explored here. Our results suggest that EFN is a potential route of exposure of neonicotinoids to beneficial insects and that further field-based studies are warranted.

The lipidome of endometrial fluid differs between implantative and non-implantative IVF cycles

OBJECTIVE

To characterize the most relevant changes in the lipidome of endometrial fluid aspirate (EFA) in non-implantative cycles.

DESIGN

Lipidomics in a prospective cohort study.

SETTINGS

Reproductive unit of a university hospital.

PATIENTS

Twenty-nine women undergoing an IVF cycle. Fifteen achieved pregnancy and 14 did not.

INTERVENTION

Endometrial fluid aspiration immediately before performing embryo transfer.

MAIN OUTCOME MEASURES

Clinical pregnancy rate and lipidomic profiles obtained on an ultra-high performance liquid chromatography coupled to time-of-flight mass spectrometry (UHPLC-ToF-MS)-based analytical platform.

RESULTS

The comparative analysis of the lipidomic patterns of endometrial fluid in implantative and non-implantative IVF cycles revealed eight altered metabolites: seven glycerophospholipids and an omega-6 polyunsaturated fatty acid. Then, women with a non-implantative cycle were accurately classified with a support vector machine algorithm including these eight lipid metabolites. The diagnostic performances of the algorithm showed an area under the receiver operating characteristic curve, sensitivity, specificity, and accuracy of 0.893 ± 0.07, 85.7%, 80.0%, and 82.8%, respectively.

CONCLUSION

A predictive lipidomic signature linked to the implantative status of the endometrial fluid has been found.

The vaginal metabolome and microbiota of cervical HPV-positive and HPV-negative women: a cross-sectional analysis

OBJECTIVE

Characterise the vaginal metabolome of cervical HPV-infected and uninfected women.

DESIGN

Cross-sectional.

SETTING

The Center for Health Behavior Research at the University of Maryland School of Public Health.

SAMPLE

Thirty-nine participants, 13 categorised as HPV-negative and 26 as HPV-positive (any genotype; HPV⁺ ), 14 of whom were positive with at least one high-risk HPV strain (hrHPV).

METHOD

Self-collected mid-vaginal swabs were profiled for bacterial composition by 16S rRNA gene amplicon sequencing, metabolites by both gas and liquid chromatography mass spectrometry, and 37 types of HPV DNA.

MAIN OUTCOME MEASURES

Metabolite abundances.

RESULTS

Vaginal microbiota clustered into Community State Type (CST) I (Lactobacillus crispatus-dominated), CST III (Lactobacillus iners-dominated), and CST IV (low-Lactobacillus, 'molecular-BV'). HPV⁺ women had higher biogenic amine and phospholipid concentrations compared with HPV^- women after adjustment for CST and cigarette smoking. Metabolomic profiles of HPV⁺ and HPV^- women differed in strata of CST. In CST III, there were higher concentrations of biogenic amines and glycogen-related metabolites in HPV⁺ women than in HPV^- women. In CST IV, there were lower concentrations of glutathione, glycogen, and phospholipid-related metabolites in HPV⁺ participants than in HPV^- participants. Across all CSTs, women with hrHPV strains had lower concentrations of amino acids, lipids, and peptides compared with women who had only low-risk HPV (lrHPV).

CONCLUSIONS

The vaginal metabolome of HPV⁺ women differed from HPV^- women in terms of several metabolites, including biogenic amines, glutathione, and lipid-related metabolites. If the temporal relation between increased levels of reduced glutathione and oxidised glutathione and HPV incidence/persistence is confirmed in future studies, anti-oxidant therapies may be considered as a non-surgical HPV control intervention.

TWEETABLE ABSTRACT

Metabolomics study: Vaginal microenvironment of HPV⁺ women may be informative for non-surgical interventions.

Integrated targeted metabolomic and lipidomic analysis: A novel approach to classifying early cystic precursors to invasive pancreatic cancer

Pancreatic cystic neoplasms (PCNs) are a highly prevalent disease of the pancreas. Among PCNs, Intraductal Papillary Mucinous Neoplasms (IPMNs) are common lesions that may progress from low-grade dysplasia (LGD) through high-grade dysplasia (HGD) to invasive cancer. Accurate discrimination of IPMN-associated neoplastic grade is an unmet clinical need. Targeted (semi)quantitative analysis of 100 metabolites and >1000 lipid species were performed on peri-operative pancreatic cyst fluid and pre-operative plasma from IPMN and serous cystic neoplasm (SCN) patients in a pancreas resection cohort (n = 35). Profiles were correlated against histological diagnosis and clinical parameters after correction for confounding factors. Integrated data modeling was used for group classification and selection of the best explanatory molecules. Over 1000 different compounds were identified in plasma and cyst fluid. IPMN profiles showed significant lipid pathway alterations compared to SCN. Integrated data modeling discriminated between IPMN and SCN with 100% accuracy and distinguished IPMN LGD or IPMN HGD and invasive cancer with up to 90.06% accuracy. Free fatty acids, ceramides, and triacylglycerol classes in plasma correlated with circulating levels of CA19-9, albumin and bilirubin. Integrated metabolomic and lipidomic analysis of plasma or cyst fluid can improve discrimination of IPMN from SCN and within PMNs predict the grade of dysplasia.

"-Omics" workflow for paleolimnological and geological archives: A review

"-Omics" is a powerful screening method with applications in molecular biology, toxicology, wildlife biology, natural product discovery, and many other fields. Genomics, proteomics, metabolomics, and lipidomics are common examples included under the "-omics" umbrella. This screening method uses combinations of untargeted, semi-targeted, and targeted analyses paired with data mining to facilitate researchers' understanding of the genome, proteins, and small organic molecules in biological systems. Recently, however, the use of "-omics" has expanded into the fields of geology, specifically petrology, and paleolimnology. Specifically, untargeted analyses stand to transform these fields as petroleomics, and sediment-"omics" become more prevalent. "-Omics" facilitates the visualization of small molecule profiles from environmental matrices (i.e. oil and sediment). Small molecule profiles can provide improved understanding of small molecules distributions throughout the environment, and how those compositions can change depending on conditions (i.e. climate change, weathering, etc.). "-Omics" also facilities discovery of next-generation biomarkers that can be used for oil source identification and as proxies for reconstructing past environmental changes. Untargeted analyses paired with data mining and multivariate statistical analyses represents a powerful suite of tools for hypothesis generation, and new method development for environmental reconstructions. Here we present an introduction to "-omics" methodology, technical terms, and examples of applications to paleolimnology and petrology. The purpose of this review is to highlight the important considerations at each step in the "-omics" workflow to produce high quality and statistically powerful data for petrological and paleolimnological applications.

Inflammatory responses relate to distinct bronchoalveolar lavage lipidome in community-acquired pneumonia patients: a pilot study

Background

Community-acquired pneumonia (CAP) is a leading cause of morbidity and mortality worldwide. Antibiotics are losing their effectiveness due to the emerging infectious diseases, the scarcity of novel antibiotics, and the contributions of antibiotic misuse and overuse to resistance. Characterization of the lipidomic response to pneumonia and exploring the “lipidomic phenotype” can provide new insight into the underlying mechanisms of pathogenesis and potential avenues for diagnostic and therapeutic treatments.

Methods

Lipid profiles of bronchoalveolar lavage fluid (BALF) samples were generated through untargeted lipidomic profiling analysis using high-performance liquid chromatography with mass spectrometry (HPLC-MS). Principal component analysis (PCA) was applied to identify possible sources of variations among samples. Partitioning clustering analysis (k-means) was employed to evaluate the existence of distinct lipidomic clusters.

Results

PCA showed that BALF lipidomes differed significantly between CAP (n = 52) and controls (n = 68, including 35 healthy volunteers and 33 patients with non-infectious lung diseases); while no clear separation was found between severe CAP and non-severe CAP cases. Lactosylceramides were the most prominently elevated lipid constituent in CAP. Clustering analysis revealed three separate lipid profiles; subjects in each cluster exhibited significant differences in disease severity, incidence of hypoxemia, percentages of phagocytes in BALF, and serum concentrations of albumin and total cholesterol (all p < 0.05). In addition, SM (d34:1) was negatively related to macrophage (adjusted r = − 0.462, p < 0.0001) and PE (18:1p/20:4) was positively correlated with polymorphonuclear neutrophil (PMN) percentages of BALF (adjusted r = 0.541, p < 0.0001). The 30-day mortality did not differ amongst three clusters (p < 0.05).

Conclusions

Our data suggest that specific lower airway lipid composition is related to different intensities of host inflammatory responses, and may contribute to functionally relevant shifts in disease pathogenesis in CAP individuals. These findings argue for the need to tailor therapy based on specific lipid profiles and related inflammatory status.

Trial registration

ClinicalTrials.gov (NCT03093220). Registered on 28 March 2017 (retrospectively registered).

Electronic supplementary material

The online version of this article (10.1186/s12931-019-1028-8) contains supplementary material, which is available to authorized users.

NMR-Based Μetabolomics of the Lipid Fraction of Organic and Conventional Bovine Milk

Origin and quality identification in dairy products is an important issue and also an extremely challenging and complex experimental procedure. The objective of the present work was to compare the metabolite profile of the lipid fraction of organic and conventional bovine milk using NMR metabolomics analysis. ¹H-NMR and 1D TOCSY NMR methods of analysis were performed on extracted lipid fraction of lyophilized milk. For this purpose, 14 organic and 16 conventional retail milk samples were collected monthly, and 64 bulk-tank (58 conventional and 6 organics) milk samples were collected over a 14-month longitudinal study in Cyprus. Data were treated with multivariate methods (PCA, PLS-DA). Minor components were identified and quantified, and modification of the currently used equations is proposed. A significantly increased % content of conjugated (9-cis, 11-trans)18:2 linoleic acid (CLA), α-linolenic acid, linoleic acid, allylic protons and total unsaturated fatty acids (UFA) and decreased % content for caproleic acid were observed in the organic samples compared to the conventional ones. The present work confirms that lipid profile is affected by contrasting management system (organic vs. conventional), and supports the potential of NMR-based metabolomics for the rapid analysis and authentication of the milk from its lipid profile.

Mass spectrometry-based metabolomics of volatiles as a new tool for understanding aroma and flavour chemistry in processed food products

BACKGROUND

When foods are processed or cooked, many chemical reactions occur involving a wide range of metabolites including sugars, amino acids and lipids. These chemical processes often lead to the formation of volatile aroma compounds that can make food tastier or may introduce off-flavours. Metabolomics tools are only now being used to study the formation of these flavour compounds in order to understand better the beneficial and less beneficial aspects of food processing.

AIM OF REVIEW

To provide a critical overview of the diverse MS-based studies carried out in recent years in food metabolomics and to review some biochemical properties and flavour characteristics of the different groups of aroma-related metabolites. A description of volatiles from processed foods, and their relevant chemical and sensorial characteristics is provided. In addition, this review also summarizes the formation of the flavour compounds from their precursors, and the interconnections between Maillard reactions and the amino acid, lipid, and carbohydrate degradation pathways.

KEY SCIENTIFIC CONCEPTS OF REVIEW

This review provides new insights into processed ingredients and describes how metabolomics will help to enable us to produce, preserve, design and distribute higher-quality foods for health promotion and better flavour.

Metabolomic diagnostics and human digital image

The existing clinical laboratory practice has limitations in terms of specificity and sensitivity of diagnosis, making the introduction of new methods in medicine more topical. Application of 'omics' technologies, especially metabolomics, allows overcoming these limitations. The composition of blood metabolites reflects the physical state of an organism at the molecular level. The analysis of blood metabolome can serve as effective means of diagnosis, implementation of which in healthcare is timely and relevant. This paper demonstrates the versatility of metabolomic diagnostics, its applicability to various diseases. We discussed the standard of human digital image, which includes the metabolomic data sufficient to make an accurate assessment of general health and carry out precision diagnostics of a wide range of diseases.

Robust volcano plot: identification of differential metabolites in the presence of outliers

BACKGROUND

The identification of differential metabolites in metabolomics is still a big challenge and plays a prominent role in metabolomics data analyses. Metabolomics datasets often contain outliers because of analytical, experimental, and biological ambiguity, but the currently available differential metabolite identification techniques are sensitive to outliers.

RESULTS

We propose a kernel weight based outlier-robust volcano plot for identifying differential metabolites from noisy metabolomics datasets. Two numerical experiments are used to evaluate the performance of the proposed technique against nine existing techniques, including the t-test and the Kruskal-Wallis test. Artificially generated data with outliers reveal that the proposed method results in a lower misclassification error rate and a greater area under the receiver operating characteristic curve compared with existing methods. An experimentally measured breast cancer dataset to which outliers were artificially added reveals that our proposed method produces only two non-overlapping differential metabolites whereas the other nine methods produced between seven and 57 non-overlapping differential metabolites.

CONCLUSION

Our data analyses show that the performance of the proposed differential metabolite identification technique is better than that of existing methods. Thus, the proposed method can contribute to analysis of metabolomics data with outliers. The R package and user manual of the proposed method are available at https://github.com/nishithkumarpaul/Rvolcano .

Missing Value Imputation Approach for Mass Spectrometry-based Metabolomics Data

Missing values exist widely in mass-spectrometry (MS) based metabolomics data. Various methods have been applied for handling missing values, but the selection can significantly affect following data analyses. Typically, there are three types of missing values, missing not at random (MNAR), missing at random (MAR), and missing completely at random (MCAR). Our study comprehensively compared eight imputation methods (zero, half minimum (HM), mean, median, random forest (RF), singular value decomposition (SVD), k-nearest neighbors (kNN), and quantile regression imputation of left-censored data (QRILC)) for different types of missing values using four metabolomics datasets. Normalized root mean squared error (NRMSE) and NRMSE-based sum of ranks (SOR) were applied to evaluate imputation accuracy. Principal component analysis (PCA)/partial least squares (PLS)-Procrustes analysis were used to evaluate the overall sample distribution. Student's t-test followed by correlation analysis was conducted to evaluate the effects on univariate statistics. Our findings demonstrated that RF performed the best for MCAR/MAR and QRILC was the favored one for left-censored MNAR. Finally, we proposed a comprehensive strategy and developed a public-accessible web-tool for the application of missing value imputation in metabolomics ( https://metabolomics.cc.hawaii.edu/software/MetImp/ ).

Missing Value Imputation Approach for Mass Spectrometry-based Metabolomics Data

Missing values exist widely in mass-spectrometry (MS) based metabolomics data. Various methods have been applied for handling missing values, but the selection can significantly affect following data analyses. Typically, there are three types of missing values, missing not at random (MNAR), missing at random (MAR), and missing completely at random (MCAR). Our study comprehensively compared eight imputation methods (zero, half minimum (HM), mean, median, random forest (RF), singular value decomposition (SVD), k-nearest neighbors (kNN), and quantile regression imputation of left-censored data (QRILC)) for different types of missing values using four metabolomics datasets. Normalized root mean squared error (NRMSE) and NRMSE-based sum of ranks (SOR) were applied to evaluate imputation accuracy. Principal component analysis (PCA)/partial least squares (PLS)-Procrustes analysis were used to evaluate the overall sample distribution. Student's t-test followed by correlation analysis was conducted to evaluate the effects on univariate statistics. Our findings demonstrated that RF performed the best for MCAR/MAR and QRILC was the favored one for left-censored MNAR. Finally, we proposed a comprehensive strategy and developed a public-accessible web-tool for the application of missing value imputation in metabolomics ( https://metabolomics.cc.hawaii.edu/software/MetImp/ ).

Performance Evaluation and Online Realization of Data-driven Normalization Methods Used in LC/MS based Untargeted Metabolomics Analysis

In untargeted metabolomics analysis, several factors (e.g., unwanted experimental & biological variations and technical errors) may hamper the identification of differential metabolic features, which requires the data-driven normalization approaches before feature selection. So far, ≥16 normalization methods have been widely applied for processing the LC/MS based metabolomics data. However, the performance and the sample size dependence of those methods have not yet been exhaustively compared and no online tool for comparatively and comprehensively evaluating the performance of all 16 normalization methods has been provided. In this study, a comprehensive comparison on these methods was conducted. As a result, 16 methods were categorized into three groups based on their normalization performances across various sample sizes. The VSN, the Log Transformation and the PQN were identified as methods of the best normalization performance, while the Contrast consistently underperformed across all sub-datasets of different benchmark data. Moreover, an interactive web tool comprehensively evaluating the performance of 16 methods specifically for normalizing LC/MS based metabolomics data was constructed and hosted at http://server.idrb.cqu.edu.cn/MetaPre/. In summary, this study could serve as a useful guidance to the selection of suitable normalization methods in analyzing the LC/MS based metabolomics data.

A Conversation on Data Mining Strategies in LC-MS Untargeted Metabolomics: Pre-Processing and Pre-Treatment Steps

Untargeted metabolomic studies generate information-rich, high-dimensional, and complex datasets that remain challenging to handle and fully exploit. Despite the remarkable progress in the development of tools and algorithms, the "exhaustive" extraction of information from these metabolomic datasets is still a non-trivial undertaking. A conversation on data mining strategies for a maximal information extraction from metabolomic data is needed. Using a liquid chromatography-mass spectrometry (LC-MS)-based untargeted metabolomic dataset, this study explored the influence of collection parameters in the data pre-processing step, scaling and data transformation on the statistical models generated, and feature selection, thereafter. Data obtained in positive mode generated from a LC-MS-based untargeted metabolomic study (sorghum plants responding dynamically to infection by a fungal pathogen) were used. Raw data were pre-processed with MarkerLynx^TM software (Waters Corporation, Manchester, UK). Here, two parameters were varied: the intensity threshold (50-100 counts) and the mass tolerance (0.005-0.01 Da). After the pre-processing, the datasets were imported into SIMCA (Umetrics, Umea, Sweden) for more data cleaning and statistical modeling. In addition, different scaling (unit variance, Pareto, etc.) and data transformation (log and power) methods were explored. The results showed that the pre-processing parameters (or algorithms) influence the output dataset with regard to the number of defined features. Furthermore, the study demonstrates that the pre-treatment of data prior to statistical modeling affects the subspace approximation outcome: e.g., the amount of variation in X-data that the model can explain and predict. The pre-processing and pre-treatment steps subsequently influence the number of statistically significant extracted/selected features (variables). Thus, as informed by the results, to maximize the value of untargeted metabolomic data, understanding of the data structures and exploration of different algorithms and methods (at different steps of the data analysis pipeline) might be the best trade-off, currently, and possibly an epistemological imperative.

Metabolomic Profiling of Post-Mortem Brain Reveals Changes in Amino Acid and Glucose Metabolism in Mental Illness Compared with Controls

Metabolomic profiling was carried out on 53 post-mortem brain samples from subjects diagnosed with schizophrenia, depression, bipolar disorder (SDB), diabetes, and controls. Chromatography on a ZICpHILIC column was used with detection by Orbitrap mass spectrometry. Data extraction was carried out with m/z Mine 2.14 with metabolite searching against an in-house database. There was no clear discrimination between the controls and the SDB samples on the basis of a principal components analysis (PCA) model of 755 identified or putatively identified metabolites. Orthogonal partial least square discriminant analysis (OPLSDA) produced clear separation between 17 of the controls and 19 of the SDB samples (R2CUM 0.976, Q2 0.671, p-value of the cross-validated ANOVA score 0.0024). The most important metabolites producing discrimination were the lipophilic amino acids leucine/isoleucine, proline, methionine, phenylalanine, and tyrosine; the neurotransmitters GABA and NAAG and sugar metabolites sorbitol, gluconic acid, xylitol, ribitol, arabinotol, and erythritol. Eight samples from diabetic brains were analysed, six of which grouped with the SDB samples without compromising the model (R2 CUM 0.850, Q2 CUM 0.534, p-value for cross-validated ANOVA score 0.00087). There appears on the basis of this small sample set to be some commonality between metabolic perturbations resulting from diabetes and from SDB.

Non-targeted UHPLC-MS metabolomic data processing methods: a comparative investigation of normalisation, missing value imputation, transformation and scaling

INTRODUCTION

The generic metabolomics data processing workflow is constructed with a serial set of processes including peak picking, quality assurance, normalisation, missing value imputation, transformation and scaling. The combination of these processes should present the experimental data in an appropriate structure so to identify the biological changes in a valid and robust manner.

OBJECTIVES

Currently, different researchers apply different data processing methods and no assessment of the permutations applied to UHPLC-MS datasets has been published. Here we wish to define the most appropriate data processing workflow.

METHODS

We assess the influence of normalisation, missing value imputation, transformation and scaling methods on univariate and multivariate analysis of UHPLC-MS datasets acquired for different mammalian samples.

RESULTS

Our studies have shown that once data are filtered, missing values are not correlated with m/z, retention time or response. Following an exhaustive evaluation, we recommend PQN normalisation with no missing value imputation and no transformation or scaling for univariate analysis. For PCA we recommend applying PQN normalisation with Random Forest missing value imputation, glog transformation and no scaling method. For PLS-DA we recommend PQN normalisation, KNN as the missing value imputation method, generalised logarithm transformation and no scaling. These recommendations are based on searching for the biologically important metabolite features independent of their measured abundance.

CONCLUSION

The appropriate choice of normalisation, missing value imputation, transformation and scaling methods differs depending on the data analysis method and the choice of method is essential to maximise the biological derivations from UHPLC-MS datasets.