Combining Machine Learning and Metabolomics to Identify Weight Gain Biomarkers

Artificial Intelligence in Metabolomics: A Current Review

Metabolomics and artificial intelligence (AI) form a synergistic partnership. Metabolomics generates large datasets comprising hundreds to thousands of metabolites with complex relationships. AI, aiming to mimic human intelligence through computational modeling, possesses extraordinary capabilities for big data analysis. In this review, we provide a recent overview of the methodologies and applications of AI in metabolomics studies in the context of systems biology and human health. We first introduce the AI concept, history, and key algorithms for machine learning and deep learning, summarizing their strengths and weaknesses. We then discuss studies that have successfully used AI across different aspects of metabolomic analysis, including analytical detection, data preprocessing, biomarker discovery, predictive modeling, and multi-omics data integration. Lastly, we discuss the existing challenges and future perspectives in this rapidly evolving field. Despite limitations and challenges, the combination of metabolomics and AI holds great promises for revolutionary advancements in enhancing human health.

Metabolomic and Lipidomic Characterization of Meningioma Grades Using LC-HRMS and Machine Learning

Meningiomas are among the most common brain tumors that arise from the leptomeningeal cover of the brain and spinal cord and account for around 37% of all central nervous system tumors. According to the World Health Organization, meningiomas are classified into three histological subtypes: benign, atypical, and anaplastic. Sometimes, meningiomas with a histological diagnosis of benign tumors show clinical characteristics and behavior of aggressive tumors. In this study, we examined the metabolomic and lipidomic profiles of meningioma tumors, focusing on comparing low-grade and high-grade tumors and identifying potential markers that can discriminate between benign and malignant tumors. High-resolution mass spectrometry coupled to liquid chromatography was used for untargeted metabolomics and lipidomics analyses of 85 tumor biopsy samples with different meningioma grades. We then applied feature selection and machine learning techniques to find the features with the highest information to aid in the diagnosis of meningioma grades. Three biomarkers were identified to differentiate low- and high-grade meningioma brain tumors. The use of mass-spectrometry-based metabolomics and lipidomics combined with machine learning analyses to prospect and characterize biomarkers associated with meningioma grades may pave the way for elucidating potential therapeutic and prognostic targets.

Targeted plasma metabolomics combined with machine learning for the diagnosis of severe acute respiratory syndrome virus type 2

Introduction

The routine clinical diagnosis of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is largely restricted to real-time reverse transcription quantitative PCR (RT-qPCR), and tests that detect SARS-CoV-2 nucleocapsid antigen. Given the diagnostic delay and suboptimal sensitivity associated with these respective methods, alternative diagnostic strategies are needed for acute infection.

Methods

We studied the use of a clinically validated liquid chromatography triple quadrupole method (LC/MS–MS) for detection of amino acids from plasma specimens. We applied machine learning models to distinguish between SARS-CoV-2-positive and negative samples and analyzed amino acid feature importance.

Results

A total of 200 samples were tested, including 70 from individuals with COVID-19, and 130 from negative controls. The top performing model overall allowed discrimination between SARS-CoV-2-positive and negative control samples with an area under the receiver operating characteristic curve (AUC) of 0.96 (95%CI 0.91, 1.00), overall sensitivity of 0.99 (95%CI 0.92, 1.00), and specificity of 0.92 (95%CI 0.85, 0.95).

Discussion

This approach holds potential as an alternative to existing methods for the rapid and accurate diagnosis of acute SARS-CoV-2 infection.

A machine learning model for visualization and dynamic clinical prediction of stroke recurrence in acute ischemic stroke patients: A real-world retrospective study

Background and purpose

Recurrent stroke accounts for 25–30% of all preventable strokes, and this study was conducted to establish a machine learning-based clinical predictive rice idol for predicting stroke recurrence within 1 year in patients with acute ischemic stroke (AIS).

Methods

A total of 645 AIS patients at The Second Affiliated Hospital of Xuzhou Medical University were screened, included and followed up for 1 year for comprehensive clinical data. Univariate and multivariate logistic regression (LR) were used to screen the risk factors of stroke recurrence. The data set was randomly divided into training set and test set according to the ratio of 7:3, and the following six prediction models were established by machine algorithm: random forest (RF), Naive Bayes model (NBC), decision tree (DT), extreme gradient boosting (XGB), gradient boosting machine (GBM) and LR. The model with the strongest prediction performance was selected by 10-fold cross-validation and receiver operating characteristic (ROC) curves, and the models were investigated for interpretability by SHAP. Finally, the models were constructed to be visualized using a web calculator.

Results

Logistic regression analysis showed that right hemisphere, homocysteine (HCY), C-reactive protein (CRP), and stroke severity (SS) were independent risk factors for the development of stroke recurrence in AIS patients. In 10-fold cross-validation, area under curve (AUC) ranked from 0.777 to 0.959. In ROC curve analysis, AUC ranged from 0.887 to 0.946. RF model has the best ability to predict stroke recurrence, and HCY has the largest contribution to the model. A web-based calculator https://mlmedicine-re-stroke2-re-stroke2-baylee.streamlitapp.com/ has been developed accordingly.

Conclusion

This study identified four independent risk factors affecting recurrence within 1 year in stroke patients, and the constructed RF-based prediction model had good performance.

Predicting metabolomic profiles from microbial composition through neural ordinary differential equations

Characterizing the metabolic profile of a microbial community is crucial for understanding its biological function and its impact on the host or environment. Metabolomics experiments directly measuring these profiles are difficult and expensive, while sequencing methods quantifying the species composition of microbial communities are well-developed and relatively cost-effective. Computational methods that are capable of predicting metabolomic profiles from microbial compositions can save considerable efforts needed for metabolomic profiling experimentally. Yet, despite existing efforts, we still lack a computational method with high prediction power, general applicability, and great interpretability. Here we develop a method - mNODE (Metabolomic profile predictor using Neural Ordinary Differential Equations), based on a state-of-the-art family of deep neural network models. We show compelling evidence that mNODE outperforms existing methods in predicting the metabolomic profiles of human microbiomes and several environmental microbiomes. Moreover, in the case of human gut microbiomes, mNODE can naturally incorporate dietary information to further enhance the prediction of metabolomic profiles. Besides, susceptibility analysis of mNODE enables us to reveal microbe-metabolite interactions, which can be validated using both synthetic and real data. The presented results demonstrate that mNODE is a powerful tool to investigate the microbiome-diet-metabolome relationship, facilitating future research on precision nutrition.

The inflammatory proteome, obesity, and medical weight loss and regain in humans

OBJECTIVE

Weight regain occurs after medical weight loss via mechanisms of post-weight-loss "metabolic adaptation." The relationship of inflammatory proteins with weight loss/regain was studied to determine a role for inflammation in metabolic adaptation.

METHODS

Seventy-four proteins central to inflammation and immune regulation (Olink) were analyzed in plasma from up to 490 participants in a trial of medical weight-loss maintenance. Cross-sectional and longitudinal associations of proteins with weight were measured using linear and mixed effects regression models and t testing, with replication in the Framingham Heart Study.

RESULTS

Broad changes in the inflammatory proteome were observed among the study cohort (60% women, 35% African American) with initial weight loss of ≈8 kg from a median 94 kg at study entry (33/74 proteins; 7 increased; 26 decreased), many of which tracked with weight regain of median ≈2 kg over the next 30 months. Ten proteins were associated with different rates of weight regain, some specifying pathways of chemotaxis and innate immune responses. Several of the observed protein associations were also linked to prevalent obesity in the Framingham Heart Study.

CONCLUSIONS

Broad changes in the inflammatory proteome track with changes in weight and may identify specific pathways that modify patterns of weight regain.

Metabolite patterns associated with individual response to supervised exercise therapy in patients with intermittent claudication

Objective

Supervised exercise therapy (SET) is the first line treatment for intermittent claudication owing to peripheral arterial disease. Despite multiple randomized controlled trials proving the efficacy of SET, there are large differences in individual patient's responses. We used plasma metabolomics to identify potential metabolic influences on the individual response to SET.

Methods

Primary metabolites, complex lipids, and lipid mediators were measured on plasma samples taken at before and after Gardner graded treadmill walking tests that were administered before and after 12 weeks of SET. We used an ensemble modeling approach to identify metabolites or changes in metabolites at specific time points that associated with interindividual variability in the functional response to SET. Specific time points analyzed included baseline metabolite levels before SET, dynamic metabolomics changes before SET, the difference in pre- and post-SET baseline metabolomics, and the difference (pre- and post-SET) of the dynamic (pre- and post-treadmill).

Results

High levels of baseline anandamide levels pre- and post-SET were associated with a worse response to SET. Increased arachidonic acid (AA) and decreased levels of the AA precursor dihomo-γ-linolenic acid across SET were associated with a worse response to SET. Participants who were able to tolerate large increases in AA during acute exercise had longer, or better, walking times both before and after SET.

Conclusions

We identified two pathways of relevance to individual response to SET that warrant further study: anandamide synthesis may activate endocannabinoid receptors, resulting in worse treadmill test performance. SET may train patients to withstand higher levels of AA, and inflammatory signaling, resulting in longer walking times.

Clinical Relevance

This manuscript describes the use of metabolomic techniques to measure the interindividual effects of SET in patients with peripheral artery disease (PAD). We identified high levels of AEA are linked to CB1 signaling and activation of inflammatory pathways. This alters energy expenditure in myoblasts by decreasing glucose uptake and may induce an acquired skeletal muscle myopathy. SET may also help participants tolerate increased levels of AA and inflammation produced during exercise, resulting in longer walking times. This data will enhance understanding of the pathophysiology of PAD and the mechanism by which SET improves walking intolerance.

Metabolomics and modelling approaches for systems metabolic engineering

Metabolic engineering involves the manipulation of microbes to produce desirable compounds through genetic engineering or synthetic biology approaches. Metabolomics involves the quantitation of intracellular and extracellular metabolites, where mass spectrometry and nuclear magnetic resonance based analytical instrumentation are often used. Here, the experimental designs, sample preparations, metabolite quenching and extraction are essential to the quantitative metabolomics workflow. The resultant metabolomics data can then be used with computational modelling approaches, such as kinetic and constraint-based modelling, to better understand underlying mechanisms and bottlenecks in the synthesis of desired compounds, thereby accelerating research through systems metabolic engineering. Constraint-based models, such as genome scale models, have been used successfully to enhance the yield of desired compounds from engineered microbes, however, unlike kinetic or dynamic models, constraint-based models do not incorporate regulatory effects. Nevertheless, the lack of time-series metabolomic data generation has hindered the usefulness of dynamic models till today. In this review, we show that improvements in automation, dynamic real-time analysis and high throughput workflows can drive the generation of more quality data for dynamic models through time-series metabolomics data generation. Spatial metabolomics also has the potential to be used as a complementary approach to conventional metabolomics, as it provides information on the localization of metabolites. However, more effort must be undertaken to identify metabolites from spatial metabolomics data derived through imaging mass spectrometry, where machine learning approaches could prove useful. On the other hand, single-cell metabolomics has also seen rapid growth, where understanding cell-cell heterogeneity can provide more insights into efficient metabolic engineering of microbes. Moving forward, with potential improvements in automation, dynamic real-time analysis, high throughput workflows, and spatial metabolomics, more data can be produced and studied using machine learning algorithms, in conjunction with dynamic models, to generate qualitative and quantitative predictions to advance metabolic engineering efforts.

Precision Medicine Approaches with Metabolomics and Artificial Intelligence

Recent technological innovations in the field of mass spectrometry have supported the use of metabolomics analysis for precision medicine. This growth has been allowed also by the application of algorithms to data analysis, including multivariate and machine learning methods, which are fundamental to managing large number of variables and samples. In the present review, we reported and discussed the application of artificial intelligence (AI) strategies for metabolomics data analysis. Particularly, we focused on widely used non-linear machine learning classifiers, such as ANN, random forest, and support vector machine (SVM) algorithms. A discussion of recent studies and research focused on disease classification, biomarker identification and early diagnosis is presented. Challenges in the implementation of metabolomics–AI systems, limitations thereof and recent tools were also discussed.

Machine Learning Data Analysis Highlights the Role of Parasutterella and Alloprevotella in Autism Spectrum Disorders

In recent years, the involvement of the gut microbiota in disease and health has been investigated by sequencing the 16S gene from fecal samples. Dysbiotic gut microbiota was also observed in Autism Spectrum Disorder (ASD), a neurodevelopmental disorder characterized by gastrointestinal symptoms. However, despite the relevant number of studies, it is still difficult to identify a typical dysbiotic profile in ASD patients. The discrepancies among these studies are due to technical factors (i.e., experimental procedures) and external parameters (i.e., dietary habits). In this paper, we collected 959 samples from eight available projects (540 ASD and 419 Healthy Controls, HC) and reduced the observed bias among studies. Then, we applied a Machine Learning (ML) approach to create a predictor able to discriminate between ASD and HC. We tested and optimized three algorithms: Random Forest, Support Vector Machine and Gradient Boosting Machine. All three algorithms confirmed the importance of five different genera, including Parasutterella and Alloprevotella. Furthermore, our results show that ML algorithms could identify common taxonomic features by comparing datasets obtained from countries characterized by latent confounding variables.

GeneCloudOmics: A Data Analytic Cloud Platform for High-Throughput Gene Expression Analysis

Gene expression profiling techniques, such as DNA microarray and RNA-Sequencing, have provided significant impact on our understanding of biological systems. They contribute to almost all aspects of biomedical research, including studying developmental biology, host-parasite relationships, disease progression and drug effects. However, the high-throughput data generations present challenges for many wet experimentalists to analyze and take full advantage of such rich and complex data. Here we present GeneCloudOmics, an easy-to-use web server for high-throughput gene expression analysis that extends the functionality of our previous ABioTrans with several new tools, including protein datasets analysis, and a web interface. GeneCloudOmics allows both microarray and RNA-Seq data analysis with a comprehensive range of data analytics tools in one package that no other current standalone software or web-based tool can do. In total, GeneCloudOmics provides the user access to 23 different data analytical and bioinformatics tasks including reads normalization, scatter plots, linear/non-linear correlations, PCA, clustering (hierarchical, k-means, t-SNE, SOM), differential expression analyses, pathway enrichments, evolutionary analyses, pathological analyses, and protein-protein interaction (PPI) identifications. Furthermore, GeneCloudOmics allows the direct import of gene expression data from the NCBI Gene Expression Omnibus database. The user can perform all tasks rapidly through an intuitive graphical user interface that overcomes the hassle of coding, installing tools/packages/libraries and dealing with operating systems compatibility and version issues, complications that make data analysis tasks challenging for biologists. Thus, GeneCloudOmics is a one-stop open-source tool for gene expression data analysis and visualization. It is freely available at http://combio-sifbi.org/GeneCloudOmics.

Lipidomic Signatures of Changes in Adiposity: A Large Prospective Study of 5849 Adults from the Australian Diabetes, Obesity and Lifestyle Study

Lipid metabolism is tightly linked to adiposity. Comprehensive lipidomic profiling offers new insights into the dysregulation of lipid metabolism in relation to weight gain. Here, we investigated the relationship of the human plasma lipidome and changes in waist circumference (WC) and body mass index (BMI). Adults (2653 men and 3196 women), 25–95 years old who attended the baseline survey of the Australian Diabetes, Obesity and Lifestyle Study (AusDiab) and the 5-year follow-up were enrolled. A targeted lipidomic approach was used to quantify 706 distinct molecular lipid species in the plasma samples. Multiple linear regression models were used to examine the relationship between the baseline lipidomic profile and changes in WC and BMI. Metabolic scores for change in WC were generated using a ridge regression model. Alkyl-diacylglycerol such as TG(O-50:2) [NL-18:1] displayed the strongest association with change in WC (β-coefficient = 0.125 cm increment per SD increment in baseline lipid level, p = 2.78 × 10⁻¹¹. Many lipid species containing linoleate (18:2) fatty acids were negatively associated with both WC and BMI gain. Compared to traditional models, multivariate models containing lipid species identify individuals at a greater risk of gaining WC: top quintile relative to bottom quintile (odds ratio, 95% CI = 5.4, 3.8–6.6 for women and 2.3, 1.7–3.0 for men). Our findings define metabolic profiles that characterize individuals at risk of weight gain or WC increase and provide important insight into the biological role of lipids in obesity.

Diagnosis Model of Hydrogen Sulfide Poisoning Based on Support Vector Machine

Introduction:

Hydrogen sulfide (H2S) is a lethal environmental and industrial poison. The mortality rate of occupational acute H2S poisoning reported in China is 23.1% ~ 50%. Due to the huge amount of information on metabolomics changes after body poisoning, it is important to use intelligent algorithms to mine multivariate interactions.

Methods:

This paper first uses GC-MS metabolomics to detect changes in the urine components of the poisoned group and control rats to form a metabolic dataset, and then uses the SVM classification algorithm in machine learning to train the hydrogen sulfide poisoning training dataset to obtain a classification recognition model. A batch of rats (n = 15) was randomly selected and exposed to 20 ppm H2S gas for 40 days (twice morning and evening, 1 hour each exposure) to prepare a chronic H2S rat poisoning model. The other rats (n = 15) were exposed to the same volume of air and 0 ppm hydrogen sulfide gas as the control group. The treated urine samples were tested using a GC-MS.

Results:

The method locates the optimal parameters of SVM, which improves the accuracy of SVM classification to 100%. This paper uses the information to gain an attribute evaluation method to screen out the top 6 biomarkers that contribute to the predicted category (Glycerol, -Hydroxybutyric acid, arabinofuranose, Pentitol, L-Tyrosine, L-Proline).

Conclusion:

The SVM diagnostic model of hydrogen sulfide poisoning constructed in this work has training time and prediction accuracy; it has achieved excellent results and provided an intelligent decision-making method for the diagnosis of hydrogen sulfide poisoning.

Using machine learning approaches for multi-omics data analysis: A review

With the development of modern high-throughput omic measurement platforms, it has become essential for biomedical studies to undertake an integrative (combined) approach to fully utilise these data to gain insights into biological systems. Data from various omics sources such as genetics, proteomics, and metabolomics can be integrated to unravel the intricate working of systems biology using machine learning-based predictive algorithms. Machine learning methods offer novel techniques to integrate and analyse the various omics data enabling the discovery of new biomarkers. These biomarkers have the potential to help in accurate disease prediction, patient stratification and delivery of precision medicine. This review paper explores different integrative machine learning methods which have been used to provide an in-depth understanding of biological systems during normal physiological functioning and in the presence of a disease. It provides insight and recommendations for interdisciplinary professionals who envisage employing machine learning skills in multi-omics studies.

Systems biology approaches integrated with artificial intelligence for optimized metabolic engineering

Metabolic engineering aims to maximize the production of bio-economically important substances (compounds, enzymes, or other proteins) through the optimization of the genetics, cellular processes and growth conditions of microorganisms. This requires detailed understanding of underlying metabolic pathways involved in the production of the targeted substances, and how the cellular processes or growth conditions are regulated by the engineering. To achieve this goal, a large system of experimental techniques, compound libraries, computational methods and data resources, including multi-omics data, are used. The recent advent of multi-omics systems biology approaches significantly impacted the field by opening new avenues to perform dynamic and large-scale analyses that deepen our knowledge on the manipulations. However, with the enormous transcriptomics, proteomics and metabolomics available, it is a daunting task to integrate the data for a more holistic understanding. Novel data mining and analytics approaches, including Artificial Intelligence (AI), can provide breakthroughs where traditional low-throughput experiment-alone methods cannot easily achieve. Here, we review the latest attempts of combining systems biology and AI in metabolic engineering research, and highlight how this alliance can help overcome the current challenges facing industrial biotechnology, especially for food-related substances and compounds using microorganisms.

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.