Genome scale metabolic network modelling for metabolic profile predictions.

Metabolic profiling (metabolomics) aims at measuring small molecules (metabolites) in complex samples like blood or urine for human health studies. While biomarker-based assessment often relies on a single molecule, metabolic profiling combines several metabolites to create a more complex and more specific fingerprint of the disease. However, in contrast to genomics, there is no unique metabolomics setup able to measure the entire metabolome. This challenge leads to tedious and resource consuming preliminary studies to be able to design the right metabolomics experiment. In that context, computer assisted metabolic profiling can be of strong added value to design metabolomics studies more quickly and efficiently. We propose a constraint-based modelling approach which predicts in silico profiles of metabolites that are more likely to be differentially abundant under a given metabolic perturbation (e.g. due to a genetic disease), using flux simulation. In genome-scale metabolic networks, the fluxes of exchange reactions, also known as the flow of metabolites through their external transport reactions, can be simulated and compared between control and disease conditions in order to calculate changes in metabolite import and export. These import/export flux differences would be expected to induce changes in circulating biofluid levels of those metabolites, which can then be interpreted as potential biomarkers or metabolites of interest. In this study, we present SAMBA (SAMpling Biomarker Analysis), an approach which simulates fluxes in exchange reactions following a metabolic perturbation using random sampling, compares the simulated flux distributions between the baseline and modulated conditions, and ranks predicted differentially exchanged metabolites as potential biomarkers for the perturbation. We show that there is a good fit between simulated metabolic exchange profiles and experimental differential metabolites detected in plasma, such as patient data from the disease database OMIM, and metabolic trait-SNP associations found in mGWAS studies. These biomarker recommendations can provide insight into the underlying mechanism or metabolic pathway perturbation lying behind observed metabolite differential abundances, and suggest new metabolites as potential avenues for further experimental analyses.

PathIntegrate: Multivariate modelling approaches for pathway-based multi-omics data integration.

As terabytes of multi-omics data are being generated, there is an ever-increasing need for methods facilitating the integration and interpretation of such data. Current multi-omics integration methods typically output lists, clusters, or subnetworks of molecules related to an outcome. Even with expert domain knowledge, discerning the biological processes involved is a time-consuming activity. Here we propose PathIntegrate, a method for integrating multi-omics datasets based on pathways, designed to exploit knowledge of biological systems and thus provide interpretable models for such studies. PathIntegrate employs single-sample pathway analysis to transform multi-omics datasets from the molecular to the pathway-level, and applies a predictive single-view or multi-view model to integrate the data. Model outputs include multi-omics pathways ranked by their contribution to the outcome prediction, the contribution of each omics layer, and the importance of each molecule in a pathway. Using semi-synthetic data we demonstrate the benefit of grouping molecules into pathways to detect signals in low signal-to-noise scenarios, as well as the ability of PathIntegrate to precisely identify important pathways at low effect sizes. Finally, using COPD and COVID-19 data we showcase how PathIntegrate enables convenient integration and interpretation of complex high-dimensional multi-omics datasets. PathIntegrate is available as an open-source Python package.

Challenges and perspectives for naming lipids in the context of lipidomics.

INTRODUCTION: Lipids are key compounds in the study of metabolism and are increasingly studied in biology projects. It is a very broad family that encompasses many compounds, and the name of the same compound may vary depending on the community where they are studied. OBJECTIVES: In addition, their structures are varied and complex, which complicates their analysis. Indeed, the structural resolution does not always allow a complete level of annotation so the actual compound analysed will vary from study to study and should be clearly stated. For all these reasons the identification and naming of lipids is complicated and very variable from one study to another, it needs to be harmonized. METHODS & RESULTS: In this position paper we will present and discuss the different way to name lipids (with chemoinformatic and semantic identifiers) and their importance to share lipidomic results. CONCLUSION: Homogenising this identification and adopting the same rules is essential to be able to share data within the community and to map data on functional networks.

Weight of evidence evaluation of the metabolism disrupting effects of triphenyl phosphate using an expert knowledge elicitation approach.

Identification of Endocrine-Disrupting Chemicals (EDCs) in a regulatory context requires a high level of evidence. However, lines of evidence (e.g. human, in vivo, in vitro or in silico) are heterogeneous and incomplete for quantifying evidence of the adverse effects and mechanisms involved. To date, for the regulatory appraisal of metabolism-disrupting chemicals (MDCs), no harmonised guidance to assess the weight of evidence has been developed at the EU or international level. To explore how to develop this, we applied a formal Expert Knowledge Elicitation (EKE) approach within the European GOLIATH project. EKE captures expert judgment in a quantitative manner and provides an estimate of uncertainty of the final opinion. As a proof of principle, we selected one suspected MDC -triphenyl phosphate (TPP) - based on its related adverse endpoints (obesity/adipogenicity) relevant to metabolic disruption and a putative Molecular Initiating Event (MIE): activation of peroxisome proliferator activated receptor gamma (PPARγ). We conducted a systematic literature review and assessed the quality of the lines of evidence with two independent groups of experts within GOLIATH, with the objective of categorising the metabolic disruption properties of TPP, by applying an EKE approach. Having followed the entire process separately, both groups arrived at the same conclusion, designating TPP as a "suspected MDC" with an overall quantitative agreement exceeding 85%, indicating robust reproducibility. The EKE method provides to be an important way to bring together scientists with diverse expertise and is recommended for future work in this area.

Lipidomic Profiling of PFOA-Exposed Mouse Liver by Multi-Modal Mass Spectrometry Analysis.

Perfluorooctanoic acid (PFOA) is a synthetic perfluorinated chemical classified as a persistent organic pollutant. PFOA has been linked to many toxic effects, including liver injury. Many studies report that PFOA exposure alters serum and hepatic lipid metabolism. However, lipidomic pathways altered by PFOA exposure are largely unknown and only a few lipid classes, mostly triacylglycerol (TG), are usually considered in lipid analysis. Here, we performed a global lipidomic analysis on the liver of PFOA-exposed (high-dose and short-duration) and control mice by combining three mass spectrometry (MS) techniques: liquid chromatography with tandem mass spectrometry (LC-MS/MS), matrix-assisted laser desorption ionization mass spectrometry imaging (MALDI-MSI), and time-of-flight secondary ion mass spectrometry (TOF-SIMS). Among all hepatic lipids identified by LC-MS/MS analysis, more than 350 were statistically impacted (increased or decreased levels) after PFOA exposure, as confirmed by multi-variate data analysis. The levels of many lipid species from different lipid classes, most notably phosphatidylethanolamine (PE), phosphatidylcholine (PC), and TG, were significantly altered. Subsequent lipidomic analysis highlights the pathways significantly impacted by PFOA exposure, with the glycerophospholipid metabolism being the most impacted, and the changes in the lipidome network, which connects all the lipid species together. MALDI-MSI displays the heterogeneous distribution of the affected lipids and PFOA, revealing different areas of lipid expression linked to PFOA localization. TOF-SIMS localizes PFOA at the cellular level, supporting MALDI-MSI results. This multi-modal MS analysis unveils the lipidomic impact of PFOA in the mouse liver after high-dose and short-term exposure and opens new opportunities in toxicology.

Compared with Milk Protein, a Wheat and Pea Protein Blend Reduces High-Fat, High-Sucrose Induced Metabolic Dysregulations while Similarly Supporting Tissue Protein Anabolism in Rats.

BACKGROUND: Plant proteins (PPs) have been associated with better cardiovascular health than animal proteins (APs) in epidemiological studies. However, the underlying metabolic mechanisms remain mostly unknown. OBJECTIVES: Using a combination of cutting-edge isotopic methods, we aimed to better characterize the differences in protein and energy metabolisms induced by dietary protein sources (PP compared with AP) in a prudent or western dietary context. METHODS: Male Wistar rats (n = 44, 8 wk old) were fed for 4.5 mo with isoproteic diets differing in their protein isolate sources, either AP (100% milk) or PP (50%:50% pea: wheat) and being normal (NFS) or high (HFS) in sucrose (6% or 15% kcal) and saturated fat (7% or 20% kcal), respectively. We measured body weight and composition, hepatic enzyme activities and lipid content, and plasma metabolites. In the intestine, liver, adipose tissues, and skeletal muscles, we concomitantly assessed the extent of amino acid (AA) trafficking using a 15N natural abundance method, the rates of macronutrient routing to dispensable AA using a 13C natural abundance method, and the metabolic fluxes of protein synthesis (PS) and de novo lipogenesis using a 2H labeling method. Data were analyzed using ANOVA and Mixed models. RESULTS: At the whole-body level, PP limited HFS-induced insulin resistance (-27% in HOMA-IR between HFS groups, P < 0.05). In the liver, PP induced lower lipid content (-17%, P < 0.01) and de novo lipogenesis (-24%, P < 0.05). In the different tissues studied, PP induced higher AA transamination accompanied by higher routings of dietary carbohydrates and lipids toward dispensable AA synthesis by glycolysis and ß-oxidation, resulting in similar tissue PS and protein mass. CONCLUSIONS: In growing rats, compared with AP, a balanced blend of PP similarly supports protein anabolism while better limiting whole-body and tissue metabolic dysregulations through mechanisms related to their less optimal AA profile for direct channeling to PS.

DEXOM: Diversity-based enumeration of optimal context-specific metabolic networks.

The correct identification of metabolic activity in tissues or cells under different conditions can be extremely elusive due to mechanisms such as post-transcriptional modification of enzymes or different rates in protein degradation, making difficult to perform predictions on the basis of gene expression alone. Context-specific metabolic network reconstruction can overcome some of these limitations by leveraging the integration of multi-omics data into genome-scale metabolic networks (GSMN). Using the experimental information, context-specific models are reconstructed by extracting from the generic GSMN the sub-network most consistent with the data, subject to biochemical constraints. One advantage is that these context-specific models have more predictive power since they are tailored to the specific tissue, cell or condition, containing only the reactions predicted to be active in such context. However, an important limitation is that there are usually many different sub-networks that optimally fit the experimental data. This set of optimal networks represent alternative explanations of the possible metabolic state. Ignoring the set of possible solutions reduces the ability to obtain relevant information about the metabolism and may bias the interpretation of the true metabolic states. In this work we formalize the problem of enumerating optimal metabolic networks and we introduce DEXOM, an unified approach for diversity-based enumeration of context-specific metabolic networks. We developed different strategies for this purpose and we performed an exhaustive analysis using simulated and real data. In order to analyze the extent to which these results are biologically meaningful, we used the alternative solutions obtained with the different methods to measure: 1) the improvement of in silico predictions of essential genes in Saccharomyces cerevisiae using ensembles of metabolic network; and 2) the detection of alternative enriched pathways in different human cancer cell lines. We also provide DEXOM as an open-source library compatible with COBRA Toolbox 3.0, available at https://github.com/MetExplore/dexom.

Pathway analysis in metabolomics: Recommendations for the use of over-representation analysis.

Over-representation analysis (ORA) is one of the commonest pathway analysis approaches used for the functional interpretation of metabolomics datasets. Despite the widespread use of ORA in metabolomics, the community lacks guidelines detailing its best-practice use. Many factors have a pronounced impact on the results, but to date their effects have received little systematic attention. Using five publicly available datasets, we demonstrated that changes in parameters such as the background set, differential metabolite selection methods, and pathway database used can result in profoundly different ORA results. The use of a non-assay-specific background set, for example, resulted in large numbers of false-positive pathways. Pathway database choice, evaluated using three of the most popular metabolic pathway databases (KEGG, Reactome, and BioCyc), led to vastly different results in both the number and function of significantly enriched pathways. Factors that are specific to metabolomics data, such as the reliability of compound identification and the chemical bias of different analytical platforms also impacted ORA results. Simulated metabolite misidentification rates as low as 4% resulted in both gain of false-positive pathways and loss of truly significant pathways across all datasets. Our results have several practical implications for ORA users, as well as those using alternative pathway analysis methods. We offer a set of recommendations for the use of ORA in metabolomics, alongside a set of minimal reporting guidelines, as a first step towards the standardisation of pathway analysis in metabolomics.

Improving lipid mapping in Genome Scale Metabolic Networks using ontologies.

INTRODUCTION: To interpret metabolomic and lipidomic profiles, it is necessary to identify the metabolic reactions that connect the measured molecules. This can be achieved by putting them in the context of genome-scale metabolic network reconstructions. However, mapping experimentally measured molecules onto metabolic networks is challenging due to differences in identifiers and level of annotation between data and metabolic networks, especially for lipids. OBJECTIVES: To help linking lipids from lipidomics datasets with lipids in metabolic networks, we developed a new matching method based on the ChEBI ontology. The implementation is freely available as a python library and in MetExplore webserver. METHODS: Our matching method is more flexible than an exact identifier-based correspondence since it allows establishing a link between molecules even if a different level of precision is provided in the dataset and in the metabolic network. For instance, it can associate a generic class of lipids present in the network with the molecular species detailed in the lipidomics dataset. This mapping is based on the computation of a distance between molecules in ChEBI ontology. RESULTS: We applied our method to a chemical library (968 lipids) and an experimental dataset (32 modulated lipids) and showed that using ontology-based mapping improves and facilitates the link with genome scale metabolic networks. Beyond network mapping, the results provide ways for improvements in terms of network curation and lipidomics data annotation. CONCLUSION: This new method being generic, it can be applied to any metabolomics data and therefore improve our comprehension of metabolic modulations.

MetExplore: collaborative edition and exploration of metabolic networks.

Metabolism of an organism is composed of hundreds to thousands of interconnected biochemical reactions responding to environmental or genetic constraints. This metabolic network provides a rich knowledge to contextualize omics data and to elaborate hypotheses on metabolic modulations. Nevertheless, performing this kind of integrative analysis is challenging for end users with not sufficiently advanced computer skills since it requires the use of various tools and web servers. MetExplore offers an all-in-one online solution composed of interactive tools for metabolic network curation, network exploration and omics data analysis. In particular, it is possible to curate and annotate metabolic networks in a collaborative environment. The network exploration is also facilitated in MetExplore by a system of interactive tables connected to a powerful network visualization module. Finally, the contextualization of metabolic elements in the network and the calculation of over-representation statistics make it possible to interpret any kind of omics data. MetExplore is a sustainable project maintained since 2010 freely available at https://metexplore.toulouse.inra.fr/metexplore2/.

Large-Scale Modeling Approach Reveals Functional Metabolic Shifts during Hepatic Differentiation.

Being able to explore the metabolism of broad metabolizing cells is of critical importance in many research fields. This article presents an original modeling solution combining metabolic network and omics data to identify modulated metabolic pathways and changes in metabolic functions occurring during differentiation of a human hepatic cell line (HepaRG). Our results confirm the activation of hepato-specific functionalities and newly evidence modulation of other metabolic pathways, which could not be evidenced from transcriptomic data alone. Our method takes advantage of the network structure to detect changes in metabolic pathways that do not have gene annotations and exploits flux analyses techniques to identify activated metabolic functions. Compared to the usual cell-specific metabolic network reconstruction approaches, it limits false predictions by considering several possible network configurations to represent one phenotype rather than one arbitrarily selected network. Our approach significantly enhances the comprehensive and functional assessment of cell metabolism, opening further perspectives to investigate metabolic shifts occurring within various biological contexts.

MetExploreViz: web component for interactive metabolic network visualization.

SUMMARY: MetExploreViz is an open source web component that can be easily embedded in any web site. It provides features dedicated to the visualization of metabolic networks and pathways and thus offers a flexible solution to analyse omics data in a biochemical context. AVAILABILITY AND IMPLEMENTATION: Documentation and link to GIT code repository (GPL 3.0 license) are available at this URL: http://metexplore.toulouse.inra.fr/metexploreViz/doc/.

Early changes in tissue amino acid metabolism and nutrient routing in rats fed a high-fat diet: evidence from natural isotope abundances of nitrogen and carbon in tissue proteins.

Little is known about how diet-induced obesity and insulin resistance affect protein and amino acid (AA) metabolism in tissues. The natural relative abundances of the heavy stable isotopes of C (δ 13C) and N (δ 15N) in tissue proteins offer novel and promising biomarkers of AA metabolism. They, respectively, reflect the use of dietary macronutrients for tissue AA synthesis and the relative metabolic use of tissue AA for oxidation v. protein synthesis. In this study, δ 13C and δ 15N were measured in the proteins of various tissues in young adult rats exposed perinatally and/or fed after weaning with a normal- or a high-fat (HF) diet, the aim being to characterise HF-induced tissue-specific changes in AA metabolism. HF feeding was shown to increase the routing of dietary fat to all tissue proteins via non-indispensable AA synthesis, but did not affect AA allocation between catabolic and anabolic processes in most tissues. However, the proportion of AA directed towards oxidation rather than protein synthesis was increased in the small intestine and decreased in the tibialis anterior muscle and adipose tissue. In adipose tissue, the AA reallocation was observed in the case of perinatal or post-weaning exposure to HF, whereas in the small intestine and tibialis anterior muscle the AA reallocation was only observed after HF exposure that covered both the perinatal and post-weaning periods. In conclusion, HF exposure induced an early reorganisation of AA metabolism involving tissue-specific effects, and in particular a decrease in the relative allocation of AA to oxidation in several peripheral tissues.

Integrated transcriptomics and metabolomics reveal signatures of lipid metabolism dysregulation in HepaRG liver cells exposed to PCB 126.

Chemical pollutant exposure is a risk factor contributing to the growing epidemic of non-alcoholic fatty liver disease (NAFLD) affecting human populations that consume a western diet. Although it is recognized that intoxication by chemical pollutants can lead to NAFLD, there is limited information available regarding the mechanism by which typical environmental levels of exposure can contribute to the onset of this disease. Here, we describe the alterations in gene expression profiles and metabolite levels in the human HepaRG liver cell line, a validated model for cellular steatosis, exposed to the polychlorinated biphenyl (PCB) 126, one of the most potent chemical pollutants that can induce NAFLD. Sparse partial least squares classification of the molecular profiles revealed that exposure to PCB 126 provoked a decrease in polyunsaturated fatty acids as well as an increase in sphingolipid levels, concomitant with a decrease in the activity of genes involved in lipid metabolism. This was associated with an increased oxidative stress reflected by marked disturbances in taurine metabolism. A gene ontology analysis showed hallmarks of an activation of the AhR receptor by dioxin-like compounds. These changes in metabolome and transcriptome profiles were observed even at the lowest concentration (100 pM) of PCB 126 tested. A decrease in docosatrienoate levels was the most sensitive biomarker. Overall, our integrated multi-omics analysis provides mechanistic insight into how this class of chemical pollutant can cause NAFLD. Our study lays the foundation for the development of molecular signatures of toxic effects of chemicals causing fatty liver diseases to move away from a chemical risk assessment based on in vivo animal experiments.

Natural isotopic signatures of variations in body nitrogen fluxes: a compartmental model analysis.

Body tissues are generally 15N-enriched over the diet, with a discrimination factor (Δ15N) that varies among tissues and individuals as a function of their nutritional and physiopathological condition. However, both 15N bioaccumulation and intra- and inter-individual Δ15N variations are still poorly understood, so that theoretical models are required to understand their underlying mechanisms. Using experimental Δ15N measurements in rats, we developed a multi-compartmental model that provides the first detailed representation of the complex functioning of the body's Δ15N system, by explicitly linking the sizes and Δ15N values of 21 nitrogen pools to the rates and isotope effects of 49 nitrogen metabolic fluxes. We have shown that (i) besides urea production, several metabolic pathways (e.g., protein synthesis, amino acid intracellular metabolism, urea recycling and intestinal absorption or secretion) are most probably associated with isotope fractionation and together contribute to 15N accumulation in tissues, (ii) the Δ15N of a tissue at steady-state is not affected by variations of its P turnover rate, but can vary according to the relative orientation of tissue free amino acids towards oxidation vs. protein synthesis, (iii) at the whole-body level, Δ15N variations result from variations in the body partitioning of nitrogen fluxes (e.g., urea production, urea recycling and amino acid exchanges), with or without changes in nitrogen balance, (iv) any deviation from the optimal amino acid intake, in terms of both quality and quantity, causes a global rise in tissue Δ15N, and (v) Δ15N variations differ between tissues depending on the metabolic changes involved, which can therefore be identified using simultaneous multi-tissue Δ15N measurements. This work provides proof of concept that Δ15N measurements constitute a new promising tool to investigate how metabolic fluxes are nutritionally or physiopathologically reorganized or altered. The existence of such natural and interpretable isotopic biomarkers promises interesting applications in nutrition and health.

Effects of dark chocolate and cocoa consumption on endothelial function and arterial stiffness in overweight adults.

The consumption of cocoa and dark chocolate is associated with a lower risk of CVD, and improvements in endothelial function may mediate this relationship. Less is known about the effects of cocoa/chocolate on the augmentation index (AI), a measure of vascular stiffness and vascular tone in the peripheral arterioles. We enrolled thirty middle-aged, overweight adults in a randomised, placebo-controlled, 4-week, cross-over study. During the active treatment (cocoa) period, the participants consumed 37 g/d of dark chocolate and a sugar-free cocoa beverage (total cocoa = 22 g/d, total flavanols (TF) = 814 mg/d). Colour-matched controls included a low-flavanol chocolate bar and a cocoa-free beverage with no added sugar (TF = 3 mg/d). Treatments were matched for total fat, saturated fat, carbohydrates and protein. The cocoa treatment significantly increased the basal diameter and peak diameter of the brachial artery by 6% (+2 mm) and basal blood flow volume by 22%. Substantial decreases in the AI, a measure of arterial stiffness, were observed in only women. Flow-mediated dilation and the reactive hyperaemia index remained unchanged. The consumption of cocoa had no effect on fasting blood measures, while the control treatment increased fasting insulin concentration and insulin resistance (P= 0·01). Fasting blood pressure (BP) remained unchanged, although the acute consumption of cocoa increased resting BP by 4 mmHg. In summary, the high-flavanol cocoa and dark chocolate treatment was associated with enhanced vasodilation in both conduit and resistance arteries and was accompanied by significant reductions in arterial stiffness in women.

PathIntegrate: Multivariate modelling approaches for pathway-based multi-omics data integration.

As terabytes of multi-omics data are being generated, there is an ever-increasing need for methods facilitating the integration and interpretation of such data. Current multi-omics integration methods typically output lists, clusters, or subnetworks of molecules related to an outcome. Even with expert domain knowledge, discerning the biological processes involved is a time-consuming activity. Here we propose PathIntegrate, a method for integrating multi-omics datasets based on pathways, designed to exploit knowledge of biological systems and thus provide interpretable models for such studies. PathIntegrate employs single-sample pathway analysis to transform multi-omics datasets from the molecular to the pathway-level, and applies a predictive single-view or multi-view model to integrate the data. Model outputs include multi-omics pathways ranked by their contribution to the outcome prediction, the contribution of each omics layer, and the importance of each molecule in a pathway. Using semi-synthetic data we demonstrate the benefit of grouping molecules into pathways to detect signals in low signal-to-noise scenarios, as well as the ability of PathIntegrate to precisely identify important pathways at low effect sizes. Finally, using COPD and COVID-19 data we showcase how PathIntegrate enables convenient integration and interpretation of complex high-dimensional multi-omics datasets. The PathIntegrate Python package is available at https://github.com/cwieder/PathIntegrate.

Isotopic and modeling investigation of long-term protein turnover in rat tissues.

Fractional synthesis rates (FSR) of tissue proteins (P) are usually measured using labeled amino acid (AA) tracer methods over short periods of time under acute, particular conditions. By combining the long-term and non-steady-state (15)N labeling of AA and P tissue fractions with compartmental modeling, we have developed a new isotopic approach to investigate the degree of compartmentation of P turnover in tissues and to estimate long-term FSR values under sustained and averaged nutritional and physiological conditions. We measured the rise-to-plateau kinetics of nitrogen isotopic enrichments (δ(15)N) in the AA and P fractions of various tissues in rats for 2 mo following a slight increase in diet δ(15)N. Using these δ(15)N kinetics and a numerical method based on a two-compartment model, we determined reliable FSR estimates for tissues in which P turnover is adequately represented by such a simple precursor-product model. This was the case for kidney, liver, plasma, and muscle, where FSR estimates were 103, 101, 58, and 11%/day, respectively. Conversely, we identified tissues, namely, skin and small intestine, where P turnover proved to be too complex to be represented by a simple two-compartment model, evidencing the higher level of subcompartmentation of the P and/or AA metabolism in these tissues. The present results support the value of this new approach in gaining cognitive and practical insights into tissue P turnover and propose new and integrated FSR values over all individual precursor AA and all diurnal variations in P kinetics.

Despite similar clinical features metabolomics reveals distinct signatures in insulin resistant and progressively obese minipigs.

Obesity is a major contributor to the silent and progressive development of type 2 diabetes (T2D) whose prevention could be improved if individuals at risk were identified earlier. Our aim is to identify early phenotypes that precede T2D in diet-induced obese minipigs. We fed four groups of minipigs (n = 5-10) either normal-fat or high-fat high-sugar diet during 2, 4, or 6 months. Morphometric features were recorded, and metabolomics and clinical parameters were assessed on fasting plasma samples. Multivariate statistical analysis on 46 morphometrical and clinical parameters allowed to differentiate 4 distinct phenotypes: NFC (control group) and three others (HF2M, HF4M, HF6M) corresponding to the different stages of the obesity progression. Compared to NFC, we observed a rapid progression of body weight and fat mass (4-, 7-, and tenfold) in obese phenotypes. Insulin resistance (IR; 2.5-fold increase of HOMA-IR) and mild dyslipidemia (1.2- and twofold increase in total cholesterol and HDL) were already present in the HF2M and remained stable in HF4M and HF6M. Plasma metabolome revealed subtle changes of 23 metabolites among the obese groups, including a progressive switch in energy metabolism from amino acids to lipids, and a transient increase in de novo lipogenesis and TCA-related metabolites in HF2M. Low anti-oxidative capacities and anti-inflammatory response metabolites were found in the HF4M, and a perturbed hexose metabolism was observed in HF6M. Overall, we show that IR and progressively obese minipigs reveal phenotype-specific metabolomic signatures for which some of the identified metabolites could be considered as potential biomarkers of early progression to TD2.