A comparative evaluation of the generalised predictive ability of eight machine learning algorithms across ten clinical metabolomics data sets for binary classification

Machine learning methods in near infrared spectroscopy for predicting sensory traits in sweetpotatoes

It has been established that near infrared (NIR) spectroscopy has the potential of estimating sensory traits given the direct spectral responses that these properties have in the NIR region. In sweetpotato, sensory and texture traits are key for improving acceptability of the crop for food security and nutrition. Studies have statistically modelled the levels of NIR spectroscopy sensory characteristics using partial least squares (PLS) regression methods. To improve prediction accuracy, there are many advanced techniques, which could enhance modelling of fresh (wet and un-processed) samples or nonlinear dependence relationships. Performance of different quantitative prediction models for sensory traits developed using different machine learning methods were compared. Overall, results show that linear methods; linear support vector machine (L-SVM), principal component regression (PCR) and PLS exhibited higher mean R2 values than other statistical methods. For all the 27 sensory traits, calibration models using L-SVM and PCR has slightly higher overall R2 (x¯ = 0.33) compared to PLS (x¯ = 0.32) and radial-based SVM (NL-SVM; x¯= 0.30). The levels of orange color intensity were the best predicted by all the calibration models (R2 = 0.87 - 0.89). The elastic net linear regression (ENR) and tree-based methods; extreme gradient boost (XGBoost) and random forest (RF) performed worse than would be expected but could possibly be improved with increased sample size. Lower average R2 values were observed for calibration models of ENR (x¯ = 0.26), XGBoost (x¯ = 0.26) and RF (x¯ = 0.22). The overall RMSE in calibration models was lower in PCR models (X = 0.82) compared to L-SVM (x¯ = 0.86) and PLS (x¯ = 0.90). ENR, XGBoost and RF also had higher RMSE (x¯ = 0.90 - 0.92). Effective wavelengths selection using the interval partial least-squares regression (iPLS), improved the performance of the models but did not perform as good as the PLS. SNV pre-treatment was useful in improving model performance.

Novel COVID-19 biomarkers identified through multi-omics data analysis: N-acetyl-4-O-acetylneuraminic acid, N-acetyl-L-alanine, N-acetyltriptophan, palmitoylcarnitine, and glycerol 1-myristate

This study aims to apply machine learning models to identify new biomarkers associated with the early diagnosis and prognosis of SARS-CoV-2 infection.Plasma and serum samples from COVID-19 patients (mild, moderate, and severe), patients with other pneumonia (but with negative COVID-19 RT-PCR), and healthy volunteers (control) from hospitals in four different countries (China, Spain, France, and Italy) were analyzed by GC-MS, LC-MS, and NMR. Machine learning models (PCA and PLS-DA) were developed to predict the diagnosis and prognosis of COVID-19 and identify biomarkers associated with these outcomes.A total of 1410 patient samples were analyzed. The PLS-DA model presented a diagnostic and prognostic accuracy of around 95% of all analyzed data. A total of 23 biomarkers (e.g., spermidine, taurine, L-aspartic, L-glutamic, L-phenylalanine and xanthine, ornithine, and ribothimidine) have been identified as being associated with the diagnosis and prognosis of COVID-19. Additionally, we also identified for the first time five new biomarkers (N-Acetyl-4-O-acetylneuraminic acid, N-Acetyl-L-Alanine, N-Acetyltriptophan, palmitoylcarnitine, and glycerol 1-myristate) that are also associated with the severity and diagnosis of COVID-19. These five new biomarkers were elevated in severe COVID-19 patients compared to patients with mild disease or healthy volunteers.The PLS-DA model was able to predict the diagnosis and prognosis of COVID-19 around 95%. Additionally, our investigation pinpointed five novel potential biomarkers linked to the diagnosis and prognosis of COVID-19: N-Acetyl-4-O-acetylneuraminic acid, N-Acetyl-L-Alanine, N-Acetyltriptophan, palmitoylcarnitine, and glycerol 1-myristate. These biomarkers exhibited heightened levels in severe COVID-19 patients compared to those with mild COVID-19 or healthy volunteers.

Machine Learning-Driven SERS Nanoendoscopy and Optophysiology

A frontier of analytical sciences is centered on the continuous measurement of molecules in or near cells, tissues, or organs, within the biological context in situ, where the molecular-level information is indicative of health status, therapeutic efficacy, and fundamental biochemical function of the host. Following the completion of the Human Genome Project, current research aims to link genes to functions of an organism and investigate how the environment modulates functional properties of organisms. New analytical methods have been developed to detect chemical changes with high spatial and temporal resolution, including minimally invasive surface-enhanced Raman scattering (SERS) nanofibers using the principles of endoscopy (SERS nanoendoscopy) or optical physiology (SERS optophysiology). Given the large spectral data sets generated from these experiments, SERS nanoendoscopy and optophysiology benefit from advances in data science and machine learning to extract chemical information from complex vibrational spectra measured by SERS. This review highlights new opportunities for intracellular, extracellular, and in vivo chemical measurements arising from the combination of SERS nanosensing and machine learning.

A Support Vector Machine-Assisted Metabolomics Approach for Non-Targeted Screening of Multi-Class Pesticides and Veterinary Drugs in Maize

The contamination risks of plant-derived foods due to the co-existence of pesticides and veterinary drugs (P&VDs) have not been fully understood. With an increasing number of unexpected P&VDs illegally added to foods, it is essential to develop a non-targeted screening method for P&VDs for their comprehensive risk assessment. In this study, a modified support vector machine (SVM)-assisted metabolomics approach by screening eligible variables to represent marker compounds of 124 multi-class P&VDs in maize was developed based on the results of high-performance liquid chromatography-tandem mass spectrometry. Principal component analysis and orthogonal partial least squares discriminant analysis indicate the existence of variables with obvious inter-group differences, which were further investigated by S-plot plots, permutation tests, and variable importance in projection to obtain eligible variables. Meanwhile, SVM recursive feature elimination under the radial basis function was employed to obtain the weight-squared values of all the variables ranging from large to small for the screening of eligible variables as well. Pairwise t-tests and fold changes of concentration were further employed to confirm these eligible variables to represent marker compounds. The results indicate that 120 out of 124 P&VDs can be identified by the SVM-assisted metabolomics method, while only 109 P&VDs can be found by the metabolomics method alone, implying that SVM can promote the screening accuracy of the metabolomics method. In addition, the method's practicability was validated by the real contaminated maize samples, which provide a bright application prospect in non-targeted screening of contaminants. The limits of detection for 120 P&VDs in maize samples were calculated to be 0.3~1.5 µg/kg.

Automated Machine Learning and Explainable AI (AutoML-XAI) for Metabolomics: Improving Cancer Diagnostics

Metabolomics generates complex data necessitating advanced computational methods for generating biological insight. While machine learning (ML) is promising, the challenges of selecting the best algorithms and tuning hyperparameters, particularly for nonexperts, remain. Automated machine learning (AutoML) can streamline this process; however, the issue of interpretability could persist. This research introduces a unified pipeline that combines AutoML with explainable AI (XAI) techniques to optimize metabolomics analysis. We tested our approach on two data sets: renal cell carcinoma (RCC) urine metabolomics and ovarian cancer (OC) serum metabolomics. AutoML, using Auto-sklearn, surpassed standalone ML algorithms like SVM and k-Nearest Neighbors in differentiating between RCC and healthy controls, as well as OC patients and those with other gynecological cancers. The effectiveness of Auto-sklearn is highlighted by its AUC scores of 0.97 for RCC and 0.85 for OC, obtained from the unseen test sets. Importantly, on most of the metrics considered, Auto-sklearn demonstrated a better classification performance, leveraging a mix of algorithms and ensemble techniques. Shapley Additive Explanations (SHAP) provided a global ranking of feature importance, identifying dibutylamine and ganglioside GM(d34:1) as the top discriminative metabolites for RCC and OC, respectively. Waterfall plots offered local explanations by illustrating the influence of each metabolite on individual predictions. Dependence plots spotlighted metabolite interactions, such as the connection between hippuric acid and one of its derivatives in RCC, and between GM3(d34:1) and GM3(18:1_16:0) in OC, hinting at potential mechanistic relationships. Through decision plots, a detailed error analysis was conducted, contrasting feature importance for correctly versus incorrectly classified samples. In essence, our pipeline emphasizes the importance of harmonizing AutoML and XAI, facilitating both simplified ML application and improved interpretability in metabolomics data science.

Python workflow for the selection and identification of marker peptides-proof-of-principle study with heated milk

The analysis of almost holistic food profiles has developed considerably over the last years. This has also led to larger amounts of data and the ability to obtain more information about health-beneficial and adverse constituents in food than ever before. Especially in the field of proteomics, software is used for evaluation, and these do not provide specific approaches for unique monitoring questions. An additional and more comprehensive way of evaluation can be done with the programming language Python. It offers broad possibilities by a large ecosystem for mass spectrometric data analysis, but needs to be tailored for specific sets of features, the research questions behind. It also offers the applicability of various machine-learning approaches. The aim of the present study was to develop an algorithm for selecting and identifying potential marker peptides from mass spectrometric data. The workflow is divided into three steps: (I) feature engineering, (II) chemometric data analysis, and (III) feature identification. The first step is the transformation of the mass spectrometric data into a structure, which enables the application of existing data analysis packages in Python. The second step is the data analysis for selecting single features. These features are further processed in the third step, which is the feature identification. The data used exemplarily in this proof-of-principle approach was from a study on the influence of a heat treatment on the milk proteome/peptidome.

Adjusting for covariates and assessing modeling fitness in machine learning using MUVR2

Motivation

Machine learning (ML) methods are frequently used in Omics research to examine associations between molecular data and for example exposures and health conditions. ML is also used for feature selection to facilitate biological interpretation. Our previous MUVR algorithm was shown to generate predictions and variable selections at state-of-the-art performance. However, a general framework for assessing modeling fitness is still lacking. In addition, enabling to adjust for covariates is a highly desired, but largely lacking trait in ML. We aimed to address these issues in the new MUVR2 framework.

Results

The MUVR2 algorithm was developed to include the regularized regression framework elastic net in addition to partial least squares and random forest modeling. Compared with other cross-validation strategies, MUVR2 consistently showed state-of-the-art performance, including variable selection, while minimizing overfitting. Testing on simulated and real-world data, we also showed that MUVR2 allows for the adjustment for covariates using elastic net modeling, but not using partial least squares or random forest.

Availability and implementation

Algorithms, data, scripts, and a tutorial are open source under GPL-3 license and available in the MUVR2 R package at https://github.com/MetaboComp/MUVR2.

Metabolomics analysis reveals novel serum metabolite alterations in cancer cachexia

Background

Cachexia is a body wasting syndrome that significantly affects well-being and prognosis of cancer patients, without effective treatment. Serum metabolites take part in pathophysiological processes of cancer cachexia, but apart from altered levels of select serum metabolites, little is known on the global changes of the overall serum metabolome, which represents a functional readout of the whole-body metabolic state. Here, we aimed to comprehensively characterize serum metabolite alterations and analyze associated pathways in cachectic cancer patients to gain new insights that could help instruct strategies for novel interventions of greater clinical benefit.

Methods

Serum was sampled from 120 metastatic cancer patients (stage UICC IV). Patients were grouped as cachectic or non-cachectic according to the criteria for cancer cachexia agreed upon international consensus (main criterium: weight loss adjusted to body mass index). Samples were pooled by cachexia phenotype and assayed using non-targeted gas chromatography-mass spectrometry (GC-MS). Normalized metabolite levels were compared using t-test (p < 0.05, adjusted for false discovery rate) and partial least squares discriminant analysis (PLS-DA). Machine-learning models were applied to identify metabolite signatures for separating cachexia states. Significant metabolites underwent MetaboAnalyst 5.0 pathway analysis.

Results

Comparative analyses included 78 cachectic and 42 non-cachectic patients. Cachectic patients exhibited 19 annotable, significantly elevated (including glucose and fructose) or decreased (mostly amino acids) metabolites associating with aminoacyl-tRNA, glutathione and amino acid metabolism pathways. PLS-DA showed distinct clusters (accuracy: 85.6%), and machine-learning models identified metabolic signatures for separating cachectic states (accuracy: 83.2%; area under ROC: 88.0%). We newly identified altered blood levels of erythronic acid and glucuronic acid in human cancer cachexia, potentially linked to pentose-phosphate and detoxification pathways.

Conclusion

We found both known and yet unknown serum metabolite and metabolic pathway alterations in cachectic cancer patients that collectively support a whole-body metabolic state with impaired detoxification capability, altered glucose and fructose metabolism, and substrate supply for increased and/or distinct metabolic needs of cachexia-associated tumors. These findings together imply vulnerabilities, dependencies and targets for novel interventions that have potential to make a significant impact on future research in an important field of cancer patient care.

Data-Driven Compound Identification in Atmospheric Mass Spectrometry

Aerosol particles found in the atmosphere affect the climate and worsen air quality. To mitigate these adverse impacts, aerosol particle formation and aerosol chemistry in the atmosphere need to be better mapped out and understood. Currently, mass spectrometry is the single most important analytical technique in atmospheric chemistry and is used to track and identify compounds and processes. Large amounts of data are collected in each measurement of current time-of-flight and orbitrap mass spectrometers using modern rapid data acquisition practices. However, compound identification remains a major bottleneck during data analysis due to lacking reference libraries and analysis tools. Data-driven compound identification approaches could alleviate the problem, yet remain rare to non-existent in atmospheric science. In this perspective, the authors review the current state of data-driven compound identification with mass spectrometry in atmospheric science and discuss current challenges and possible future steps toward a digital era for atmospheric mass spectrometry.

Prediction of Clinical Remission with Adalimumab Therapy in Patients with Ulcerative Colitis by Fourier Transform-Infrared Spectroscopy Coupled with Machine Learning Algorithms

We aimed to develop prediction models for clinical remission associated with adalimumab treatment in patients with ulcerative colitis (UC) using Fourier transform-infrared (FT-IR) spectroscopy coupled with machine learning (ML) algorithms. This prospective, observational, multicenter study enrolled 62 UC patients and 30 healthy controls. The patients were treated with adalimumab for 56 weeks, and clinical remission was evaluated using the Mayo score. Baseline fecal samples were collected and analyzed using FT-IR spectroscopy. Various data preprocessing methods were applied, and prediction models were established by 10-fold cross-validation using various ML methods. Orthogonal partial least squares-discriminant analysis (OPLS-DA) showed a clear separation of healthy controls and UC patients, applying area normalization and Pareto scaling. OPLS-DA models predicting short- and long-term remission (8 and 56 weeks) yielded area-under-the-curve values of 0.76 and 0.75, respectively. Logistic regression and a nonlinear support vector machine were selected as the best prediction models for short- and long-term remission, respectively (accuracy of 0.99). In external validation, prediction models for short-term (logistic regression) and long-term (decision tree) remission performed well, with accuracy values of 0.73 and 0.82, respectively. This was the first study to develop prediction models for clinical remission associated with adalimumab treatment in UC patients by fecal analysis using FT-IR spectroscopy coupled with ML algorithms. Logistic regression, nonlinear support vector machines, and decision tree were suggested as the optimal prediction models for remission, and these were noninvasive, simple, inexpensive, and fast analyses that could be applied to personalized treatments.

Insufficient sleep and weekend recovery sleep: classification by a metabolomics-based machine learning ensemble

Although weekend recovery sleep is common, the physiological responses to weekend recovery sleep are not fully elucidated. Identifying molecular biomarkers that represent adequate versus insufficient sleep could help advance our understanding of weekend recovery sleep. Here, we identified potential molecular biomarkers of insufficient sleep and defined the impact of weekend recovery sleep on these biomarkers using metabolomics in a randomized controlled trial. Healthy adults (n = 34) were randomized into three groups: control (CON: 9-h sleep opportunities); sleep restriction (SR: 5-h sleep opportunities); or weekend recovery (WR: simulated workweek of 5-h sleep opportunities followed by ad libitum weekend recovery sleep and then 2 days with 5-h sleep opportunities). Blood for metabolomics was collected on the simulated Monday immediately following the weekend. Nine machine learning models, including a machine learning ensemble, were built to classify samples from SR versus CON. Notably, SR showed decreased glycerophospholipids and sphingolipids versus CON. The machine learning ensemble showed the highest G-mean performance and classified 50% of the WR samples as insufficient sleep. Our findings show insufficient sleep and recovery sleep influence the plasma metabolome and suggest more than one weekend of recovery sleep may be necessary for the identified biomarkers to return to healthy adequate sleep levels.

Sex Determination of Human Nails Based on Attenuated Total Reflection Fourier Transform Infrared Spectroscopy in Forensic Context

This study reports on the successful use of a machine learning approach using attenuated total reflectance Fourier transform infrared (ATR FT-IR) spectroscopy for the classification and prediction of a donor's sex from the fingernails of 63 individuals. A significant advantage of ATR FT-IR is its ability to provide a specific spectral signature for different samples based on their biochemical composition. The infrared spectrum reveals unique vibrational features of a sample based on the different absorption frequencies of the individual functional groups. This technique is fast, simple, non-destructive, and requires only small quantities of measured material with minimal-to-no sample preparation. However, advanced multivariate techniques are needed to elucidate multiplex spectral information and the small differences caused by donor characteristics. We developed an analytical method using ATR FT-IR spectroscopy advanced with machine learning (ML) based on 63 donors' fingernails (37 males, 26 females). The PLS-DA and ANN models were established, and their generalization abilities were compared. Here, the PLS scores from the PLS-DA model were used for an artificial neural network (ANN) to create a classification model. The proposed ANN model showed a greater potential for predictions, and it was validated against an independent dataset, which resulted in 92% correctly classified spectra. The results of the study are quite impressive, with 100% accuracy achieved in correctly classifying donors as either male or female at the donor level. Here, we underscore the potential of ML algorithms to leverage the selectivity of ATR FT-IR spectroscopy and produce predictions along with information about the level of certainty in a scientifically defensible manner. This proof-of-concept study demonstrates the value of ATR FT-IR spectroscopy as a forensic tool to discriminate between male and female donors, which is significant for forensic applications.

Automated machine learning and explainable AI (AutoML-XAI) for metabolomics: improving cancer diagnostics

Motivation

Metabolomics generates complex data necessitating advanced computational methods for generating biological insight. While machine learning (ML) is promising, the challenges of selecting the best algorithms and tuning hyperparameters, particularly for non-experts, remain. Automated machine learning (AutoML) can streamline this process; however, the issue of interpretability could persist. This research introduces a unified pipeline that combines AutoML with explainable AI (XAI) techniques to optimize metabolomics analysis.

Results

We tested our approach on two datasets: renal cell carcinoma (RCC) urine metabolomics and ovarian cancer (OC) serum metabolomics. AutoML, using auto-sklearn, surpassed standalone ML algorithms such as SVM and random forest in differentiating between RCC and healthy controls, as well as OC patients and those with other gynecological cancers (Non-OC). Auto-sklearn employed a mix of algorithms and ensemble techniques, yielding a superior performance (AUC of 0.97 for RCC and 0.85 for OC). Shapley Additive Explanations (SHAP) provided a global ranking of feature importance, identifying dibutylamine and ganglioside GM(d34:1) as the top discriminative metabolites for RCC and OC, respectively. Waterfall plots offered local explanations by illustrating the influence of each metabolite on individual predictions. Dependence plots spotlighted metabolite interactions, such as the connection between hippuric acid and one of its derivatives in RCC, and between GM3(d34:1) and GM3(18:1_16:0) in OC, hinting at potential mechanistic relationships. Through decision plots, a detailed error analysis was conducted, contrasting feature importance for correctly versus incorrectly classified samples. In essence, our pipeline emphasizes the importance of harmonizing AutoML and XAI, facilitating both simplified ML application and improved interpretability in metabolomics data science.

Availability

https://github.com/obifarin/automl-xai-metabolomics.

Multivariate prediction of Saliva Precipitation Index for relating selected chemical parameters of red wines to the sensory perception of astringency

Astringency is an essential sensory attribute of red wine closely related to the saliva precipitation upon contact with the wine. In this study a data matrix of 52 physico-chemical parameters was used to predict the Saliva Precipitation Index (SPI) in 110 Italian mono-varietal red wines using partial least squares regression (PLSr) with variable selection by Variable Importance for Projection (VIP) and the significance of regression coefficients. The final PLSr model, evaluated using a test data set, had 3 components and yielded an R2test of 0.630 and an RMSEtest of 0.994, with 19 independent variables whose regression coefficients were all significant at p < 0.05. Variables selected in the final model according to the decreasing magnitude of their absolute regression coefficient include the following: Procyanidin B1, Epicatechin terminal unit, Total aldehydes, Protein content, Vanillin assay, 520 nm, Polysaccharide content, Epigallocatechin PHL, Tartaric acid, Volatile acidity, Titratable acidity, Catechin terminal unit, Proanthocyanidin assay, pH, Tannin-Fe/Anthocyanin, Buffer capacity, Epigallocatechin PHL gallate, Catechin + epicatechin PHL, and Tannin-Fe. These results can be used to better understand the physico-chemical relationship underlying astringency in red wine.

A critical review of machine-learning for "multi-omics" marine metabolite datasets

During the last decade, genomic, transcriptomic, proteomic, metabolomic, and other omics datasets have been generated for a wide range of marine organisms, and even more are still on the way. Marine organisms possess unique and diverse biosynthetic pathways contributing to the synthesis of novel secondary metabolites with significant bioactivities. As marine organisms have a greater tendency to adapt to stressed environmental conditions, the chance to identify novel bioactive metabolites with potential biotechnological application is very high. This review presents a comprehensive overview of the available "-omics" and "multi-omics" approaches employed for characterizing marine metabolites along with novel data integration tools. The need for the development of machine-learning algorithms for "multi-omics" approaches is briefly discussed. In addition, the challenges involved in the analysis of "multi-omics" data and recommendations for conducting "multi-omics" study were discussed.

Metabolic Differences in Diabetic Kidney Disease Patients with Normoalbuminuria versus Moderately Increased Albuminuria

Key Points

The pathophysiological mechanisms of diabetic kidney disease (DKD) with normal (nonalbuminuric DKD) versus moderately increased albuminuria (A-DKD) are not well-understood. Fatty acid biosynthesis and oxydation, gluconeogenesis, TCA cycle, and glucose-alanine cycle were more disturbed in patients with A-DKD compared with those with nonalbuminuric DKD with identical eGFR. DKD patients with and without microalbuminuria could represent different clinical phenotypes.

Background

The pathophysiological mechanisms of diabetic kidney disease (DKD) with normal versus moderately increased albuminuria (nonalbuminuric DKD [NA-DKD] and A-DKD) are currently not well-understood and could have implications for diagnosis and treatment.

Methods

Fourteen patients with NA-DKD with urine albumin–creatinine ratio <3 mg/mmol, 26 patients with A-DKD with albumin–creatinine ratio 3–29 mg/mmol, and 60 age- and sex-matched healthy controls were randomly chosen from a population-based cohort study (Nord-Trøndelag Health Study-3, Norway). Seventy-four organic acids, 21 amino acids, 21 biogenic acids, 40 acylcarnitines, 14 sphingomyelins, and 88 phosphatidylcholines were quantified in urine. One hundred forty-six patients with diabetes from the US-based Chronic Renal Insufficiency Cohort study were used to verify main findings.

Results

Patients with NA-DKD and A-DKD had similar age, kidney function, diabetes treatment, and other traditional risk factors. Still, partial least-squares discriminant analysis showed strong metabolite-based separation (R2, 0.82; Q2, 0.52), with patients with NA-DKD having a metabolic profile positioned between the profiles of healthy controls and patients with A-DKD. Seventy-five metabolites contributed significantly to separation between NA-DKD and A-DKD (variable importance in projection scores ≥1.0) with propionylcarnitine (C3), phosphatidylcholine C38:4, medium-chained (C8) fatty acid octenedioic acid, and lactic acid as the top metabolites (variable importance in projection scores, 2.7–2.2). Compared with patients with NA-DKD, those with A-DKD had higher levels of short-chained acylcarnitines, higher long-chained fatty acid levels with more double bounds, higher branched-chain amino acid levels, and lower TCA cycle intermediates. The main findings were similar by random forest analysis and in the Chronic Renal Insufficiency Cohort study. Formal enrichment analysis indicated that fatty acid biosynthesis and oxydation, gluconeogenesis, TCA cycle, and glucose-alanine cycle were more disturbed in patients with A-DKD compared with those with NA-DKD with identical eGFR. We also found indications of a Warburg-like effect in patients with A-DKD (i.e. , metabolism of glucose to lactate despite adequate oxygen).

Conclusion

DKD patients with normoalbuminuria differ substantially in their metabolic disturbances compared with patients with moderately increase albuminuria and could represent different clinical phenotypes.

Development of a Robust Consensus Modeling Approach for Identifying Cellular and Media Metabolites Predictive of Mesenchymal Stromal Cell Potency

Mesenchymal stromal cells (MSCs) have shown promise in regenerative medicine applications due in part to their ability to modulate immune cells. However, MSCs demonstrate significant functional heterogeneity in terms of their immunomodulatory function because of differences in MSC donor/tissue source, as well as non-standardized manufacturing approaches. As MSC metabolism plays a critical role in their ability to expand to therapeutic numbers ex vivo, we comprehensively profiled intracellular and extracellular metabolites throughout the expansion process to identify predictors of immunomodulatory function (T-cell modulation and indoleamine-2,3-dehydrogenase (IDO) activity). Here, we profiled media metabolites in a non-destructive manner through daily sampling and nuclear magnetic resonance (NMR), as well as MSC intracellular metabolites at the end of expansion using mass spectrometry (MS). Using a robust consensus machine learning approach, we were able to identify panels of metabolites predictive of MSC immunomodulatory function for 10 independent MSC lines. This approach consisted of identifying metabolites in 2 or more machine learning models and then building consensus models based on these consensus metabolite panels. Consensus intracellular metabolites with high predictive value included multiple lipid classes (such as phosphatidylcholines, phosphatidylethanolamines, and sphingomyelins) while consensus media metabolites included proline, phenylalanine, and pyruvate. Pathway enrichment identified metabolic pathways significantly associated with MSC function such as sphingolipid signaling and metabolism, arginine and proline metabolism, and autophagy. Overall, this work establishes a generalizable framework for identifying consensus predictive metabolites that predict MSC function, as well as guiding future MSC manufacturing efforts through identification of high-potency MSC lines and metabolic engineering.

A comparison of different machine-learning techniques for the selection of a panel of metabolites allowing early detection of brain tumors

Metabolomics combined with machine learning methods (MLMs), is a powerful tool for searching novel diagnostic panels. This study was intended to use targeted plasma metabolomics and advanced MLMs to develop strategies for diagnosing brain tumors. Measurement of 188 metabolites was performed on plasma samples collected from 95 patients with gliomas (grade I-IV), 70 with meningioma, and 71 healthy individuals as a control group. Four predictive models to diagnose glioma were prepared using 10 MLMs and a conventional approach. Based on the cross-validation results of the created models, the F1-scores were calculated, then obtained values were compared. Subsequently, the best algorithm was applied to perform five comparisons involving gliomas, meningiomas, and controls. The best results were obtained using the newly developed hybrid evolutionary heterogeneous decision tree (EvoHDTree) algorithm, which was validated using Leave-One-Out Cross-Validation, resulting in an F1-score for all comparisons in the range of 0.476-0.948 and the area under the ROC curves ranging from 0.660 to 0.873. Brain tumor diagnostic panels were constructed with unique metabolites, which reduces the likelihood of misdiagnosis. This study proposes a novel interdisciplinary method for brain tumor diagnosis based on metabolomics and EvoHDTree, exhibiting significant predictive coefficients.

Targeted metabolomic profiling as a tool for diagnostics of patients with non-small-cell lung cancer

Lung cancer is referred to as the second most common cancer worldwide and is mainly associated with complex diagnostics and the absence of personalized therapy. Metabolomics may provide significant insights into the improvement of lung cancer diagnostics through identification of the specific biomarkers or biomarker panels that characterize the pathological state of the patient. We performed targeted metabolomic profiling of plasma samples from individuals with non-small cell lung cancer (NSLC, n = 100) and individuals without any cancer or chronic pathologies (n = 100) to identify the relationship between plasma endogenous metabolites and NSLC by means of modern comprehensive bioinformatics tools, including univariate analysis, multivariate analysis, partial correlation network analysis and machine learning. Through the comparison of metabolomic profiles of patients with NSCLC and noncancer individuals, we identified significant alterations in the concentration levels of metabolites mainly related to tryptophan metabolism, the TCA cycle, the urea cycle and lipid metabolism. Additionally, partial correlation network analysis revealed new ratios of the metabolites that significantly distinguished the considered groups of participants. Using the identified significantly altered metabolites and their ratios, we developed a machine learning classification model with an ROC AUC value equal to 0.96. The developed machine learning lung cancer model may serve as a prototype of the approach for the in-time diagnostics of lung cancer that in the future may be introduced in routine clinical use. Overall, we have demonstrated that the combination of metabolomics and up-to-date bioinformatics can be used as a potential tool for proper diagnostics of patients with NSCLC.

Spatial and temporal resolution of metabolic dysregulation in the Sugen hypoxia model of pulmonary hypertension

Although PAH is partially attributed to disordered metabolism, previous human studies have mostly examined circulating metabolites at a single time point, potentially overlooking crucial disease biology. Current knowledge gaps include an understanding of temporal changes that occur within and across relevant tissues, and whether observed metabolic changes might contribute to disease pathobiology. We utilized targeted tissue metabolomics in the Sugen hypoxia (SuHx) rodent model to investigate tissue-specific metabolic relationships with pulmonary hypertensive features over time using regression modeling and time-series analysis. Our hypotheses were that some metabolic changes would precede phenotypic changes, and that examining metabolic interactions across heart, lung, and liver tissues would yield insight into interconnected metabolic mechanisms. To support the relevance of our findings, we sought to establish links between SuHx tissue metabolomics and human PAH -omics data using bioinformatic predictions. Metabolic differences between and within tissue types were evident by Day 7 postinduction, demonstrating distinct tissue-specific metabolism in experimental pulmonary hypertension. Various metabolites demonstrated significant tissue-specific associations with hemodynamics and RV remodeling. Individual metabolite profiles were dynamic, and some metabolic shifts temporally preceded the emergence of overt pulmonary hypertension and RV remodeling. Metabolic interactions were observed such that abundance of several liver metabolites modulated lung and RV metabolite-phenotype relationships. Taken all together, regression analyses, pathway analyses and time-series analyses implicated aspartate and glutamate signaling and transport, glycine homeostasis, lung nucleotide abundance, and oxidative stress as relevant to early PAH pathobiology. These findings offer valuable insights into potential targets for early intervention in PAH.

Metabolic Fingerprints of Effective Fluoxetine Treatment in the Prefrontal Cortex of Chronically Socially Isolated Rats: Marker Candidates and Predictive Metabolites

The increasing prevalence of depression requires more effective therapy and the understanding of antidepressants' mode of action. We carried out untargeted metabolomics of the prefrontal cortex of rats exposed to chronic social isolation (CSIS), a rat model of depression, and/or fluoxetine treatment using liquid chromatography-high resolution mass spectrometry. The behavioral phenotype was assessed by the forced swim test. To analyze the metabolomics data, we employed univariate and multivariate analysis and biomarker capacity assessment using the receiver operating characteristic (ROC) curve. We also identified the most predictive biomarkers using a support vector machine with linear kernel (SVM-LK). Upregulated myo-inositol following CSIS may represent a potential marker of depressive phenotype. Effective fluoxetine treatment reversed depressive-like behavior and increased sedoheptulose 7-phosphate, hypotaurine, and acetyl-L-carnitine contents, which were identified as marker candidates for fluoxetine efficacy. ROC analysis revealed 4 significant marker candidates for CSIS group discrimination, and 10 for fluoxetine efficacy. SVM-LK with accuracies of 61.50% or 93.30% identified a panel of 7 or 25 predictive metabolites for depressive-like behavior or fluoxetine effectiveness, respectively. Overall, metabolic fingerprints combined with the ROC curve and SVM-LK may represent a new approach to identifying marker candidates or predictive metabolites for ongoing disease or disease risk and treatment outcome.

Using Multivariate Data Analysis to Project Performance in Biathletes and Cross-Country Skiers

PURPOSE

To determine whether competitive performance, as defined by International Biathlon Union (IBU) and International Ski Federation (FIS) points in biathlon and cross-country (XC) skiing, respectively, can be projected using a combination of anthropometric and physiological metrics. Shooting accuracy was also included in the biathlon models.

METHODS

Data were analyzed using multivariate methods from 45 (23 female and 22 male) biathletes and 202 (86 female and 116 male) XC skiers who were all members of senior national teams, national development teams, or ski-university or high school invite-only programs (age range: 16-36 y). Anthropometric and physiological characteristics were assessed via dual-energy X-ray absorptiometry and incremental roller-ski treadmill tests, respectively. Shooting accuracy was assessed via an outdoor standardized testing protocol.

RESULTS

Valid projective models were identified for female biathletes' IBU points (R2 = .80/Q2 = .65) and female XC skiers' FIS distance (R2 = .81/Q2 = .74) and sprint (R2 = .81/Q2 = .70) points. No valid models were identified for the men. The most important variables for the projection of IBU points were shooting accuracy, speeds at blood lactate concentrations of 4 and 2 mmol·L-1, peak aerobic power, and lean mass. The most important variables for the projection of FIS distance and sprint points were speeds at blood lactate concentrations of 4 and 2 mmol·L-1 and peak aerobic power.

CONCLUSIONS

This study highlights the relative importance of specific anthropometric, physiological, and shooting-accuracy metrics in female biathletes and XC skiers. The data can help to identify the specific metrics that should be targeted when monitoring athletes' progression and designing training plans.

A Review of Machine Learning Methods Recently Applied to FTIR Spectroscopy Data for the Analysis of Human Blood Cells

Machine learning (ML) is a broad term encompassing several methods that allow us to learn from data. These methods may permit large real-world databases to be more rapidly translated to applications to inform patient-provider decision-making. This paper presents a review of articles that discuss the use of Fourier transform infrared (FTIR) spectroscopy and ML for human blood analysis between the years 2019-2023. The literature review was conducted to identify published research of employed ML linked with FTIR for distinction between pathological and healthy human blood cells. The articles' search strategy was implemented and studies meeting the eligibility criteria were evaluated. Relevant data related to the study design, statistical methods, and strengths and limitations were identified. A total of 39 publications in the last 5 years (2019-2023) were identified and evaluated for this review. Diverse methods, statistical packages, and approaches were used across the identified studies. The most common methods included support vector machine (SVM) and principal component analysis (PCA) approaches. Most studies applied internal validation and employed more than one algorithm, while only four studies applied one ML algorithm to the data. A wide variety of approaches, algorithms, statistical software, and validation strategies were employed in the application of ML methods. There is a need to ensure that multiple ML approaches are used, the model selection strategy is clearly defined, and both internal and external validation are necessary to be sure that the discrimination of human blood cells is being made with the highest efficient evidence.

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.

Detecting Grapevine Virus Infections in Red and White Winegrape Canopies Using Proximal Hyperspectral Sensing

Grapevine virus-associated disease such as grapevine leafroll disease (GLD) affects grapevine health worldwide. Current diagnostic methods are either highly costly (laboratory-based diagnostics) or can be unreliable (visual assessments). Hyperspectral sensing technology is capable of measuring leaf reflectance spectra that can be used for the non-destructive and rapid detection of plant diseases. The present study used proximal hyperspectral sensing to detect virus infection in Pinot Noir (red-berried winegrape cultivar) and Chardonnay (white-berried winegrape cultivar) grapevines. Spectral data were collected throughout the grape growing season at six timepoints per cultivar. Partial least squares-discriminant analysis (PLS-DA) was used to build a predictive model of the presence or absence of GLD. The temporal change of canopy spectral reflectance showed that the harvest timepoint had the best prediction result. Prediction accuracies of 96% and 76% were achieved for Pinot Noir and Chardonnay, respectively. Our results provide valuable information on the optimal time for GLD detection. This hyperspectral method can also be deployed on mobile platforms including ground-based vehicles and unmanned aerial vehicles (UAV) for large-scale disease surveillance in vineyards.

Evaluation of natural ageing responses on Burmese amber durability by FTIR spectroscopy with PLSR and ANN models

Amber ageing is an inevitable process, which is very important in precious organic gemstone relics protection. In order to explore the mechanism of amber ageing and estimate the durability of Burmese amber, this research investigates the changing spectral features of Burmese ageing amber via Fourier Transform Infrared Spectroscopy (FTIR) and solid ¹³C Nuclear Magnetic Resonance spectroscopy (NMR) and develops the regression models for its micro-hardness by micro-FTIR spectra. The Partial Least Squares Regression (PLSR) and Artificial Neural Networks (ANN) methods as well as Competitive Adaptive Reweighted Sampling (CARS) algorithm for wavelength variables selection have been applied to predict and assess the Vickers hardness of amber samples with different ageing degrees. As a result, the FTIR and the solid ¹³C NMR spectra reveal that the contents of CO groups (of esters) increase substantially, and which of the other oxygenic groups (CO (of acids), COC, COCC) increase modestly in amber ageing. When comparing with the results of four different models (PLSR, ANN, CARS-PLSR and CARS-ANN), the CARS-PLSR model obtained the optimal results as follows: the squared correlation coefficient of calibration(R²cal) is 0.9230 and the root mean square error of calibration (RMSEC) is 1.2977 HV; the squared correlation coefficient of prediction (R²pre) is 0.7762 and the root mean square error of prediction (RMSEP) is 2.2208 HV. The overall results sufficiently demonstrate that FTIR spectroscopy technique coupled with appropriate chemometrics methods are very promising tools to estimate and predict the hardness property of Burmese ageing amber.

Feature Reduction for the Classification of Bruise Damage to Apple Fruit Using a Contactless FT-NIR Spectroscopy with Machine Learning

Spectroscopy data are useful for modelling biological systems such as predicting quality parameters of horticultural products. However, using the wide spectrum of wavelengths is not practical in a production setting. Such data are of high dimensional nature and they tend to result in complex models that are not easily understood. Furthermore, collinearity between different wavelengths dictates that some of the data variables are redundant and may even contribute noise. The use of variable selection methods is one efficient way to obtain an optimal model, andthis was the aim of this work. Taking advantage of a non-contact spectrometer, near infrared spectral data in the range of 800-2500 nm were used to classify bruise damage in three apple cultivars, namely 'Golden Delicious', 'Granny Smith' and 'Royal Gala'. Six prominent machine learning classification algorithms were employed, and two variable selection methods were used to determine the most relevant wavelengths for the problem of distinguishing between bruised and non-bruised fruit. The selected wavelengths clustered around 900 nm, 1300 nm, 1500 nm and 1900 nm. The best results were achieved using linear regression and support vector machine based on up to 40 wavelengths: these methods reached precision values in the range of 0.79-0.86, which were all comparable (within error bars) to a classifier based on the entire range of frequencies. The results also provided an open-source based framework that is useful towards the development of multi-spectral applications such as rapid grading of apples based on mechanical damage, and it can also be emulated and applied for other types of defects on fresh produce.

Interpretable machine learning with tree-based shapley additive explanations: Application to metabolomics datasets for binary classification

Machine learning (ML) models are used in clinical metabolomics studies most notably for biomarker discoveries, to identify metabolites that discriminate between a case and control group. To improve understanding of the underlying biomedical problem and to bolster confidence in these discoveries, model interpretability is germane. In metabolomics, partial least square discriminant analysis (PLS-DA) and its variants are widely used, partly due to the model's interpretability with the Variable Influence in Projection (VIP) scores, a global interpretable method. Herein, Tree-based Shapley Additive explanations (SHAP), an interpretable ML method grounded in game theory, was used to explain ML models with local explanation properties. In this study, ML experiments (binary classification) were conducted for three published metabolomics datasets using PLS-DA, random forests, gradient boosting, and extreme gradient boosting (XGBoost). Using one of the datasets, PLS-DA model was explained using VIP scores, while one of the best-performing models, a random forest model, was interpreted using Tree SHAP. The results show that SHAP has a more explanation depth than PLS-DA's VIP, making it a powerful method for rationalizing machine learning predictions from metabolomics studies.

Identifying Evacuation Needs and Resources Based on Volunteered Geographic Information: A Case of the Rainstorm in July 2021, Zhengzhou, China

Recently, global climate change has led to a high incidence of extreme weather and natural disasters. How to reduce its impact has become an important topic. However, the studies that both consider the disaster's real-time geographic information and environmental factors in severe rainstorms are still not enough. Volunteered geographic information (VGI) data that was generated during disasters offered possibilities for improving the emergency management abilities of decision-makers and the disaster self-rescue abilities of citizens. Through the case study of the extreme rainstorm disaster in Zhengzhou, China, in July 2021, this paper used machine learning to study VGI issued by residents. The vulnerable people and their demands were identified based on the SOS messages. The importance of various indicators was analyzed by combining open data from socio-economic and built-up environment elements. Potential safe areas with shelter resources in five administrative districts in the disaster-prone central area of Zhengzhou were identified based on these data. This study found that VGI can be a reliable data source for future disaster research. The characteristics of rainstorm hazards were concluded from the perspective of affected people and environmental indicators. The policy recommendations for disaster prevention in the context of public participation were also proposed.

Applications of machine learning in metabolomics: Disease modeling and classification

Metabolomics research has recently gained popularity because it enables the study of biological traits at the biochemical level and, as a result, can directly reveal what occurs in a cell or a tissue based on health or disease status, complementing other omics such as genomics and transcriptomics. Like other high-throughput biological experiments, metabolomics produces vast volumes of complex data. The application of machine learning (ML) to analyze data, recognize patterns, and build models is expanding across multiple fields. In the same way, ML methods are utilized for the classification, regression, or clustering of highly complex metabolomic data. This review discusses how disease modeling and diagnosis can be enhanced via deep and comprehensive metabolomic profiling using ML. We discuss the general layout of a metabolic workflow and the fundamental ML techniques used to analyze metabolomic data, including support vector machines (SVM), decision trees, random forests (RF), neural networks (NN), and deep learning (DL). Finally, we present the advantages and disadvantages of various ML methods and provide suggestions for different metabolic data analysis scenarios.

Periodontal disease metabolomics signatures from different biofluids: a systematic review

BACKGROUND

Periodontitis is resulted from a complex interaction between genetics and epigenetics, microbial factors, and the host response. Metabolomics analyses reflect both the steady-state physiological equilibrium of cells or organisms as well as their dynamic metabolic responses to environmental stimuli.

AIM OF REVIEW

This systematic review of the literature aimed to assess which low molecular weight metabolites are more often found in biological fluids of individuals with periodontitis compared to individuals with gingivitis or periodontal health.

KEY SCIENTIFIC CONCEPTS OF REVIEW

All the included studies employed untargeted analysis. One or more biological fluids were analyzed, including saliva (n = 14), gingival crevicular fluid (n = 6), mouthwash (n = 1), serum (n = 3) and plasma (n = 1). Fifty-six main metabolites related to periodontitis have been identified in at least two independent studies by NMR spectroscopy or MS-based metabolomics. Saliva was the main biological fluid sampled. It is noteworthy that 14 metabolites of the 56 detected were identified as main metabolites in all studies that sampled the saliva. The majority of metabolites found consistently among studies were amino acids, organic acids and derivates: acetate, alanine, butyrate, formate, GABA, lactate, propionate, phenylalanine and valine. They were either up- or down-regulated in the studies or this information was not mentioned. The main metabolic pathway was related to phenylalanine, tyrosine and tryptophan biosynthesis. Metabolites more frequently found in individuals with periodontitis were related to both the host and to microorganism responses. Future studies are needed, and they should follow some methodological standards to facilitate their comparison.

Interpretable machine learning methods for predictions in systems biology from omics data

Machine learning has become a powerful tool for systems biologists, from diagnosing cancer to optimizing kinetic models and predicting the state, growth dynamics, or type of a cell. Potential predictions from complex biological data sets obtained by “omics” experiments seem endless, but are often not the main objective of biological research. Often we want to understand the molecular mechanisms of a disease to develop new therapies, or we need to justify a crucial decision that is derived from a prediction. In order to gain such knowledge from data, machine learning models need to be extended. A recent trend to achieve this is to design “interpretable” models. However, the notions around interpretability are sometimes ambiguous, and a universal recipe for building well-interpretable models is missing. With this work, we want to familiarize systems biologists with the concept of model interpretability in machine learning. We consider data sets, data preparation, machine learning methods, and software tools relevant to omics research in systems biology. Finally, we try to answer the question: “What is interpretability?” We introduce views from the interpretable machine learning community and propose a scheme for categorizing studies on omics data. We then apply these tools to review and categorize recent studies where predictive machine learning models have been constructed from non-sequential omics data.

A narrative review of metabolomics in the era of "-omics": integration into clinical practice for inborn errors of metabolism

BACKGROUND AND OBJECTIVE

Traditional targeted metabolomic investigations identify a pre-defined list of analytes in samples and have been widely used for decades in the diagnosis and monitoring of inborn errors of metabolism (IEMs). Recent technological advances have resulted in the development and maturation of untargeted metabolomics: a holistic, unbiased, analytical approach to detecting metabolic disturbances in human disease. We aim to provide a summary of untargeted metabolomics [focusing on tandem mass spectrometry (MS-MS)] and its application in the field of IEMs.

METHODS

Data for this review was identified through a literature search using PubMed, Google Scholar, and personal repositories of articles collected by the authors. Findings are presented within several sections describing the metabolome, the current use of targeted metabolomics in the diagnostic pathway of patients with IEMs, the more recent integration of untargeted metabolomics into clinical care, and the limitations of this newly employed analytical technique.

KEY CONTENT AND FINDINGS

Untargeted metabolomic investigations are increasingly utilized in screening for rare disorders, improving understanding of cellular and subcellular physiology, discovering novel biomarkers, monitoring therapy, and functionally validating genomic variants. Although the untargeted metabolomic approach has some limitations, this "next generation metabolic screening" platform is becoming increasingly affordable and accessible.

CONCLUSIONS

When used in conjunction with genomics and the other promising "-omic" technologies, untargeted metabolomics has the potential to revolutionize the diagnostics of IEMs (and other rare disorders), improving both clinical and health economic outcomes.

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 based on metabolomics reveals potential targets and biomarkers for primary Sjogren’s syndrome

Background: Using machine learning based on metabolomics, this study aimed to construct an effective primary Sjogren’s syndrome (pSS) diagnostics model and reveal the potential targets and biomarkers of pSS.

Methods: From a total of 39 patients with pSS and 38 healthy controls (HCs), serum specimens were collected. The samples were analyzed by ultra-high-performance liquid chromatography coupled with high-resolution mass spectrometry. Three machine learning algorithms, including the least absolute shrinkage and selection operator (LASSO), random forest (RF), and extreme gradient boosting (XGBoost), were used to build the pSS diagnosis models. Afterward, four machine learning methods were used to reduce the dimensionality of the metabolomics data. Finally, metabolites with significant differences were screened and pathway analysis was conducted.

Results: The area under the curve (AUC), sensitivity, and specificity of LASSO, RF and XGBoost test set all reached 1.00. Orthogonal partial least squares discriminant analysis was used to classify the metabolomics data. By combining the results of the univariate false discovery rate and the importance of the variable in projection, we identified 21 significantly different metabolites. Using these 21 metabolites for diagnostic modeling, the AUC, sensitivity, and specificity of LASSO, RF, and XGBoost all reached 1.00. Metabolic pathway analysis revealed that these 21 metabolites are highly correlated with amino acid and lipid metabolisms. On the basis of 21 metabolites, we screened the important variables in the models. Further, five common variables were obtained by intersecting the important variables of three models. Based on these five common variables, the AUC, sensitivity, and specificity of LASSO, RF, and XGBoost all reached 1.00.2-Hydroxypalmitic acid, L-carnitine and cyclic AMP were found to be potential targets and specific biomarkers for pSS.

Conclusion: The combination of machine learning and metabolomics can accurately distinguish between patients with pSS and HCs. 2-Hydroxypalmitic acid, L-carnitine and cyclic AMP were potential targets and biomarkers for pSS.

Framework for Testing Robustness of Machine Learning-Based Classifiers

There has been a rapid increase in the number of artificial intelligence (AI)/machine learning (ML)-based biomarker diagnostic classifiers in recent years. However, relatively little work has focused on assessing the robustness of these biomarkers, i.e., investigating the uncertainty of the AI/ML models that these biomarkers are based upon. This paper addresses this issue by proposing a framework to evaluate the already-developed classifiers with regard to their robustness by focusing on the variability of the classifiers' performance and changes in the classifiers' parameter values using factor analysis and Monte Carlo simulations. Specifically, this work evaluates (1) the importance of a classifier's input features and (2) the variability of a classifier's output and model parameter values in response to data perturbations. Additionally, it was found that one can estimate a priori how much replacement noise a classifier can tolerate while still meeting accuracy goals. To illustrate the evaluation framework, six different AI/ML-based biomarkers are developed using commonly used techniques (linear discriminant analysis, support vector machines, random forest, partial-least squares discriminant analysis, logistic regression, and multilayer perceptron) for a metabolomics dataset involving 24 measured metabolites taken from 159 study participants. The framework was able to correctly predict which of the classifiers should be less robust than others without recomputing the classifiers itself, and this prediction was then validated in a detailed analysis.

AI/ML-driven advances in untargeted metabolomics and exposomics for biomedical applications

Metabolomics describes a high-throughput approach for measuring a repertoire of metabolites and small molecules in biological samples. One utility of untargeted metabolomics, unbiased global analysis of the metabolome, is to detect key metabolites as contributors to, or readouts of, human health and disease. In this perspective, we discuss how artificial intelligence (AI) and machine learning (ML) have promoted major advances in untargeted metabolomics workflows and facilitated pivotal findings in the areas of disease screening and diagnosis. We contextualize applications of AI and ML to the emerging field of high-resolution mass spectrometry (HRMS) exposomics, which unbiasedly detects endogenous metabolites and exogenous chemicals in human tissue to characterize exposure linked with disease outcomes. We discuss the state of the science and suggest potential opportunities for using AI and ML to improve data quality, rigor, detection, and chemical identification in untargeted metabolomics and exposomics studies.

Diagnosis and prognosis of COVID-19 employing analysis of patients' plasma and serum via LC-MS and machine learning

OBJECTIVE

To implement and evaluate machine learning (ML) algorithms for the prediction of COVID-19 diagnosis, severity, and fatality and to assess biomarkers potentially associated with these outcomes.

MATERIAL AND METHODS

Serum (n = 96) and plasma (n = 96) samples from patients with COVID-19 (acute, severe and fatal illness) from two independent hospitals in China were analyzed by LC-MS. Samples from healthy volunteers and from patients with pneumonia caused by other viruses (i.e. negative RT-PCR for COVID-19) were used as controls. Seven different ML-based models were built: PLS-DA, ANNDA, XGBoostDA, SIMCA, SVM, LREG and KNN.

RESULTS

The PLS-DA model presented the best performance for both datasets, with accuracy rates to predict the diagnosis, severity and fatality of COVID-19 of 93%, 94% and 97%, respectively. Low levels of the metabolites ribothymidine, 4-hydroxyphenylacetoylcarnitine and uridine were associated with COVID-19 positivity, whereas high levels of N-acetyl-glucosamine-1-phosphate, cysteinylglycine, methyl isobutyrate, l-ornithine and 5,6-dihydro-5-methyluracil were significantly related to greater severity and fatality from COVID-19.

CONCLUSION

The PLS-DA model can help to predict SARS-CoV-2 diagnosis, severity and fatality in daily practice. Some biomarkers typically increased in COVID-19 patients' serum or plasma (i.e. ribothymidine, N-acetyl-glucosamine-1-phosphate, l-ornithine, 5,6-dihydro-5-methyluracil) should be further evaluated as prognostic indicators of the disease.

Formal definition of the MARS method for quantifying the unique target class discoveries of selected machine classifiers

Conventional binary classification performance metrics evaluate either general measures (accuracy, F score) or specific aspects (precision, recall) of a model’s classifying ability. As such, these metrics, derived from the model’s confusion matrix, provide crucial insight regarding classifier-data interactions. However, modern- day computational capabilities have allowed for the creation of increasingly complex models that share nearly identical classification performance. While traditional performance metrics remain as essential indicators of a classifier’s individual capabilities, their ability to differentiate between models is limited. In this paper, we present the methodology for MARS (Method for Assessing Relative Sensitivity/ Specificity) ShineThrough and MARS Occlusion scores, two novel binary classification performance metrics, designed to quantify the distinctiveness of a classifier’s predictive successes and failures, relative to alternative classifiers. Being able to quantitatively express classifier uniqueness adds a novel classifier-classifier layer to the process of model evaluation and could improve ensemble model-selection decision making. By calculating both conventional performance measures, and proposed MARS metrics for a simple classifier prediction dataset, we demonstrate that the proposed metrics’ informational strengths synergize well with those of traditional metrics, delivering insight complementary to that of conventional metrics.

Fish metabolome from sub-urban lakes of the Paris area (France) and potential influence of noxious metabolites produced by cyanobacteria

The recent democratization of high-throughput molecular phenotyping allows the rapid expansion of promising untargeted multi-dimensional approaches (e.g. epigenomics, transcriptomics, proteomics, and/or metabolomics). Indeed, these emerging omics tools, processed for ecologically relevant species, may present innovative perspectives for environmental assessments, that could provide early warning of eco(toxico)logical impairments. In a previous pilot study (Sotton et al., Chemosphere 2019), we explore by ¹H NMR the bio-indicative potential of metabolomics analyses on the liver of 2 sentinel fish species (Perca fluviatilis and Lepomis gibbosus) collected in 8 water bodies of the peri-urban Paris' area (France). In the present study, we further investigate on the same samples the high potential of high-throughput UHPLC-HRMS/MS analyses. We show that the LC-MS metabolome investigation allows a clear separation of individuals according to the species, but also according to their respective sampling lakes. Interestingly, similar variations of Perca and Lepomis metabolomes occur locally indicating that site-specific environmental constraints drive the metabolome variations which seem to be influenced by the production of noxious molecules by cyanobacterial blooms in certain lakes. Thus, the development of such reliable environmental metabolomics approaches appears to constitute an innovative bio-indicative tool for the assessment of ecological stress, such as toxigenic cyanobacterial blooms, and aim at being further follow up.

Metabolomics in pulmonary medicine - extracting the most from your data

The metabolome enables unprecedented insight into biochemistry, providing an integrated signature of the genome, transcriptome, proteome and exposome. Measurement requires rigorous protocols combined with specialised data analysis to achieve its promise.https://bit.ly/3yPiYkQ

Diagnostic approach in adult-onset neurometabolic diseases

Neurometabolic diseases are a group of individually rare but numerous and heterogeneous genetic diseases best known to paediatricians. The more recently reported adult forms may present with phenotypes strikingly different from paediatric ones and may mimic other more common neurological disorders in adults. Furthermore, unlike most neurogenetic diseases, many neurometabolic diseases are treatable, with both conservative and more recent innovative therapeutics. However, the phenotypical complexity of this group of diseases and the growing number of specialised biochemical tools account for a significant diagnostic delay and underdiagnosis. We reviewed all series and case reports of patients with a confirmed neurometabolic disease and a neurological onset after the age of 10 years, with a focus on the 36 treatable ones, and classified these diseases according to their most relevant clinical manifestations. The biochemical diagnostic approach of neurometabolic diseases lays on the use of numerous tests studying a set of metabolites, an enzymatic activity or the function of a given pathway; and therapeutic options aim to restore the enzyme activity or metabolic function, limit the accumulation of toxic substrates or substitute the deficient products. A quick diagnosis of a treatable neurometabolic disease can have a major impact on patients, leading to the stabilisation of the disease and cease of repeated diagnostic investigations, and allowing for familial screening. For the aforementioned, in addition to an exhaustive and clinically meaningful review of these diseases, we propose a simplified diagnostic approach for the neurologist with the aim to help determine when to suspect a neurometabolic disease and how to proceed in a rational manner. We also discuss the place of next-generation sequencing technologies in the diagnostic process, for which deep phenotyping of patients (both clinical and biochemical) is necessary for improving their diagnostic yield.

Monosodium urate crystals regulate a unique JNK-dependent macrophage metabolic and inflammatory response

Monosodium urate crystals (MSUc) induce inflammation in vivo without prior priming, raising the possibility of an initial cell-autonomous phase. Here, using genome-wide transcriptomic analysis and biochemical assays, we demonstrate that MSUc alone induce a metabolic-inflammatory transcriptional program in non-primed human and murine macrophages that is markedly distinct to that induced by LPS. Genes uniquely upregulated in response to MSUc belong to lipid and amino acid metabolism, glycolysis, and SLC transporters. This upregulation leads to a metabolic rewiring in sera from individuals and mice with acute gouty arthritis. Mechanistically, the initiating inflammatory-metabolic changes in acute gout flares are regulated through a persistent expression and increased binding of JUN to the promoter of target genes through JNK signaling-but not P38-in a process that is different than after LPS stimulation and independent of inflammasome activation. Finally, pharmacological JNK inhibition limits MSUc-induced inflammation in animal models of acute gouty inflammation.

Multispectral Wavebands Selection for the Detection of Potential Foreign Materials in Fresh-Cut Vegetables

Ensuring the quality of fresh-cut vegetables is the greatest challenge for the food industry and is equally as important to consumers (and their health). Several investigations have proven the necessity of advanced technology for detecting foreign materials (FMs) in fresh-cut vegetables. In this study, the possibility of using near infrared spectral analysis as a potential technique was investigated to identify various types of FMs in seven common fresh-cut vegetables by selecting important wavebands. Various waveband selection methods, such as the weighted regression coefficient (WRC), variable importance in projection (VIP), sequential feature selection (SFS), successive projection algorithm (SPA), and interval PLS (iPLS), were used to investigate the optimal multispectral wavebands to classify the FMs and vegetables. The application of selected wavebands was further tested using NIR imaging, and the results showed good potentiality by identifying 99 out of 107 FMs. The results indicate the high applicability of the multispectral NIR imaging technique to detect FMs in fresh-cut vegetables for industrial application.

Reproducible untargeted metabolomics workflow for exhaustive MS2 data acquisition of MS1 features

Unknown features in untargeted metabolomics and non-targeted analysis (NTA) are identified using fragment ions from MS/MS spectra to predict the structures of the unknown compounds. The precursor ion selected for fragmentation is commonly performed using data dependent acquisition (DDA) strategies or following statistical analysis using targeted MS/MS approaches. However, the selected precursor ions from DDA only cover a biased subset of the peaks or features found in full scan data. In addition, different statistical analysis can select different precursor ions for MS/MS analysis, which make the post-hoc validation of ions selected following a secondary analysis impossible for precursor ions selected by the original statistical method. Here we propose an automated, exhaustive, statistical model-free workflow: paired mass distance-dependent analysis (PMDDA), for reproducible untargeted mass spectrometry MS2 fragment ion collection of unknown compounds found in MS1 full scan. Our workflow first removes redundant peaks from MS1 data and then exports a list of precursor ions for pseudo-targeted MS/MS analysis on independent peaks. This workflow provides comprehensive coverage of MS2 collection on unknown compounds found in full scan analysis using a “one peak for one compound” workflow without a priori redundant peak information. We compared pseudo-spectra formation and the number of MS2 spectra linked to MS1 data using the PMDDA workflow to that obtained using CAMERA and RAMclustR algorithms. More annotated compounds, molecular networks, and unique MS/MS spectra were found using PMDDA compared with CAMERA and RAMClustR. In addition, PMDDA can generate a preferred ion list for iterative DDA to enhance coverage of compounds when instruments support such functions. Finally, compounds with signals in both positive and negative modes can be identified by the PMDDA workflow, to further reduce redundancies. The whole workflow is fully reproducible as a docker image xcmsrocker with both the original data and the data processing template.

Combining MALDI-MS with machine learning for metabolomic characterization of lung cancer patient sera

An increasing amount of evidence has proven that serum metabolites can instantly reflect disease states. Therefore, sensitive and reproducible detection of serum metabolites in a high-throughput manner is urgently needed for clinical diagnosis. Matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) is a high-throughput platform for metabolite detection, but it is hindered by significant signal fluctuations because of the "sweet spot" effect of organic matrices. Here, by screening two transformation methods and four normalization techniques to reduce the significant signal fluctuations of the DHB matrix, an integrated MALDI-MS data processing approach combined with machine learning methods was established to reveal metabolic biomarkers of lung cancer. In our study, 13 distinctive features with statistically significant differences (p < 0.001) between 34 lung cancer patients and 26 healthy controls were selected as significant potential biomarkers of lung cancer. 6 out of the 13 distinctive features were identified as intact metabolites. Our results demonstrate the potential for clinical application of MALDI-MS in serum metabolomics for biomarker screening in lung cancer.

Urinary Metabolomic Changes Accompanying Albuminuria Remission following Gastric Bypass Surgery for Type 2 Diabetic Kidney Disease

In the Microvascular Outcomes after Metabolic Surgery randomised clinical trial (MOMS RCT, NCT01821508), combined metabolic surgery (gastric bypass) plus medical therapy (CSM) was superior to medical therapy alone (MTA) as a means of achieving albuminuria remission at 2-year follow-up in patients with obesity and early diabetic kidney disease (DKD). In the present study, we assessed the urinary ¹H-NMR metabolome in a subgroup of patients from both arms of the MOMS RCT at baseline and 6-month follow-up. Whilst CSM and MTA both reduced the urinary excretion of sugars, CSM generated a distinctive urinary metabolomic profile characterised by increases in host–microbial co-metabolites (N-phenylacetylglycine, trimethylamine N-oxide, and 4-aminobutyrate (GABA)) and amino acids (arginine and glutamine). Furthermore, reductions in aromatic amino acids (phenylalanine and tyrosine), as well as branched-chain amino acids (BCAAs) and related catabolites (valine, leucine, 3-hydroxyisobutyrate, 3-hydroxyisovalerate, and 3-methyl-2-oxovalerate), were observed following CSM but not MTA. Improvements in BMI did not correlate with improvements in metabolic and renal indices following CSM. Conversely, urinary metabolites changed by CSM at 6 months were moderately to strongly correlated with improvements in blood pressure, glycaemia, triglycerides, and albuminuria up to 24 months following treatment initiation, highlighting the potential involvement of these shifts in the urinary metabolomic profile in the metabolic and renoprotective effects of CSM.

Multivariate analysis of NMR-based metabolomic data

Nuclear magnetic resonance (NMR) spectroscopy allows for simultaneous detection of a wide range of metabolites and lipids. As metabolites act together in complex metabolic networks, they are often highly correlated, and optimal biological insight is achieved when using methods that take the correlation into account. For this reason, latent-variable-based methods, such as principal component analysis and partial least-squares discriminant analysis, are widely used in metabolomic studies. However, with increasing availability of larger population cohorts, and a shift from analysis of spectral data to using quantified metabolite levels, both more traditional statistical approaches and alternative machine learning methods have become more widely used. This review aims at providing an overview of the current state-of-the-art multivariate methods for the analysis of NMR-based metabolomic data as well as alternative methods, highlighting their strengths and limitations.