Application of kernel principal component analysis and computational machine learning to exploration of metabolites strongly associated with diet

Screening for obstructive sleep apnea in patients with cancer - a machine learning approach

Background

Obstructive sleep apnea (OSA) is a highly prevalent sleep disorder associated with daytime sleepiness, fatigue, and increased all-cause mortality risk in patients with cancer. Existing screening tools for OSA do not account for the interaction of cancer-related features that may increase OSA risk.

Study Design and Methods

This is a retrospective study of patients with cancer at a single tertiary cancer institution who underwent a home sleep apnea test (HSAT) to evaluate for OSA. Unsupervised machine learning (ML) was used to reduce the dimensions and extract significant features associated with OSA. ML classifiers were applied to principal components and model hyperparameters were optimized using k-fold cross-validation. Training models for OSA were subsequently tested and compared with the STOP-Bang questionnaire on a prospective unseen test set of patients who underwent an HSAT.

Results

From a training dataset of 249 patients, kernel principal component analysis (PCA) extracted eight components through dimension reduction to explain the maximum variance with OSA at 98%. Predictors of OSA were smoking, asthma, chronic kidney disease, STOP-Bang score, race, diabetes, radiation to head/neck/thorax (RT-HNT), type of cancer, and cancer metastases. Of the ML models, PCA + RF had the highest sensitivity (96.8%), specificity (92.3%), negative predictive value (92%), F1 score (0.93), and ROC-AUC score (0.88). The PCA + RF screening algorithm also performed better than the STOP-Bang questionnaire alone when tested on a prospective unseen test set.

Conclusions

The PCA + RF ML model had the highest accuracy in screening for OSA in patients with cancer. History of RT-HNT, cancer metastases, and type of cancer were identified as cancer-related risk factors for OSA.

Primary Metabolite Screening Shows Significant Differences between Embryogenic and Non-Embryogenic Callus of Tamarillo (Solanum betaceum Cav.)

Tamarillo is a solanaceous tree that has been extensively studied in terms of in vitro clonal propagation, namely somatic embryogenesis. In this work, a protocol of indirect somatic embryogenesis was applied to obtain embryogenic and non-embryogenic callus from leaf segments. Nuclear magnetic resonance spectroscopy was used to analyze the primary metabolome of these distinct calli to elucidate possible differentiation mechanisms from the common genetic background callus. Standard multivariate analysis methods were then applied, and were complemented by univariate statistical methods to identify differentially expressed primary metabolites and related metabolic pathways. The results showed carbohydrate and lipid metabolism to be the most relevant in all the calli assayed, with most discriminant metabolites being fructose, glucose and to a lesser extent choline. The glycolytic rate was higher in embryogenic calli, which shows, overall, a higher rate of sugar catabolism and a different profile of phospholipids with a choline/ethanolamine analysis. In general, our results show that a distinct primary metabolome between embryogenic and non-embryogenic calli occurs and that intracellular levels of fructose and sucrose and the glucose to sucrose ratio seem to be good candidates as biochemical biomarkers of embryogenic competence.

An agroecological structure model of compost-soil-plant interactions for sustainable organic farming

Compost is used worldwide as a soil conditioner for crops, but its functions have still been explored. Here, the omics profiles of carrots were investigated, as a root vegetable plant model, in a field amended with compost fermented with thermophilic Bacillaceae for growth and quality indices. Exposure to compost significantly increased the productivity, antioxidant activity, color, and taste of the carrot root and altered the soil bacterial composition with the levels of characteristic metabolites of the leaf, root, and soil. Based on the data, structural equation modeling (SEM) estimated that amino acids, antioxidant activity, flavonoids and/or carotenoids in plants were optimally linked by exposure to compost. The SEM of the soil estimated that the genus Paenibacillus and nitrogen compounds were optimally involved during exposure. These estimates did not show a contradiction between the whole genomic analysis of compost-derived Paenibacillus isolates and the bioactivity data, inferring the presence of a complex cascade of plant growth-promoting effects and modulation of the nitrogen cycle by the compost itself. These observations have provided information on the qualitative indicators of compost in complex soil-plant interactions and offer a new perspective for chemically independent sustainable agriculture through the efficient use of natural nitrogen.

Computational estimation of sediment symbiotic bacterial structures of seagrasses overgrowing downstream of onshore aquaculture

Coastal seagrass meadows are essential in blue carbon and aquatic ecosystem services. However, this ecosystem has suffered severe eutrophication and destruction due to the expansion of aquaculture. Therefore, methods for the flourishing of seagrass are still being explored. Here, data from 49 public coastal surveys on the distribution of seagrass and seaweed around the onshore aquaculture facilities are revalidated, and an exceptional area where the seagrass Zostera marina thrives was found near the shore downstream of the onshore aquaculture facility. To evaluate the characteristics of the sediment for growing seagrass, physicochemical properties and bacterial ecological evaluations of the sediment were conducted. Evaluation of chemical properties in seagrass sediments confirmed a significant increase in total carbon and a decrease in zinc content. Association analysis and linear discriminant analysis refined bacterial candidates specified in seagrass overgrown- and nonovergrown-sediment. Energy landscape analysis indicated that the symbiotic bacterial groups of seagrass sediment were strongly affected by the distance close to the seagrass-growing aquaculture facility despite their bacterial population appearing to fluctuate seasonally. The bacterial population there showed an apparent decrease in the pathogen candidates belonging to the order Flavobacteriales. Moreover, structure equation modeling and a linear non-Gaussian acyclic model based on the machine learning data estimated an optimal sediment symbiotic bacterial group candidate for seagrass growth as follows: the Lachnospiraceae and Ruminococcaceae families as gut-inhabitant bacteria, Rhodobacteraceae as photosynthetic bacteria, and Desulfobulbaceae as cable bacteria modulating oxygen or nitrate reduction and oxidation of sulfide. These observations confer a novel perspective on the sediment symbiotic bacterial structures critical for blue carbon and low-pathogenic marine ecosystems in aquaculture.

A potential network structure of symbiotic bacteria involved in carbon and nitrogen metabolism of wood-utilizing insect larvae

Effective biological utilization of wood biomass is necessary worldwide. Since several insect larvae can use wood biomass as a nutrient source, studies on their digestive microbial structures are expected to reveal a novel rule underlying wood biomass processing. Here, structural inferences for inhabitant bacteria involved in carbon and nitrogen metabolism for beetle larvae, an insect model, were performed to explore the potential rules. Bacterial analysis of larval feces showed enrichment of the phyla Chroloflexi, Gemmatimonadetes, and Planctomycetes, and the genera Bradyrhizobium, Chonella, Corallococcus, Gemmata, Hyphomicrobium, Lutibacterium, Paenibacillus, and Rhodoplanes, as bacteria potential involved in plant growth promotion, nitrogen cycle modulation, and/or environmental protection. The fecal abundances of these bacteria were not necessarily positively correlated with their abundances in the habitat, indicating that they were selectively enriched in the feces of the larvae. Correlation and association analyses predicted that common fecal bacteria might affect carbon and nitrogen metabolism. Based on these hypotheses, structural equation modeling (SEM) statistically estimated that inhabitant bacterial groups involved in carbon and nitrogen metabolism were composed of the phylum Gemmatimonadetes and Planctomycetes, and the genera Bradyrhizobium, Corallococcus, Gemmata, and Paenibacillus, which were among the fecal-enriched bacteria. Nevertheless, the selected common bacteria, i.e., the phyla Acidobacteria, Armatimonadetes, and Bacteroidetes and the genera Candidatus Solibacter, Devosia, Fimbriimonas, Gemmatimonas Opitutus, Sphingobium, and Methanobacterium, were necessary to obtain good fit indices in the SEM. In addition, the composition of the bacterial groups differed depending upon metabolic targets, carbon and nitrogen, and their stable isotopes, δ¹³C and δ¹⁵N, respectively. Thus, the statistically derived causal structural models highlighted that the larval fecal-enriched bacteria and common symbiotic bacteria might selectively play a role in wood biomass carbon and nitrogen metabolism. This information could confer a new perspective that helps us use wood biomass more efficiently and might stimulate innovation in environmental industries in the future.

NMR in Metabolomics: From Conventional Statistics to Machine Learning and Neural Network Approaches

NMR measurements combined with chemometrics allow achieving a great amount of information for the identification of potential biomarkers responsible for a precise metabolic pathway. These kinds of data are useful in different fields, ranging from food to biomedical fields, including health science. The investigation of the whole set of metabolites in a sample, representing its fingerprint in the considered condition, is known as metabolomics and may take advantage of different statistical tools. The new frontier is to adopt self-learning techniques to enhance clustering or classification actions that can improve the predictive power over large amounts of data. Although machine learning is already employed in metabolomics, deep learning and artificial neural networks approaches were only recently successfully applied. In this work, we give an overview of the statistical approaches underlying the wide range of opportunities that machine learning and neural networks allow to perform with accurate metabolites assignment and quantification.Various actual challenges are discussed, such as proper metabolomics, deep learning architectures and model accuracy.

Calibration Experiments of CFOSAT Wavelength in the Southern South China Sea by Artificial Neural Networks

The wave data measured by CFOSAT (China France Oceanography Satellite) have been validated mainly based on numerical model outputs and altimetry products on a global scale. It is still necessary to further calibrate the data for specific regions, e.g., the southern South China Sea. This study analyses the practicability of calibrating the dominant wavelength by using artificial neural networks and mean impact value analysis based on two sets of buoy data with a 2-year observation period and contemporaneous ERA5 reanalysis data. The artificial neural network modeling experiments are repeated 1000 times randomly by Monte Carlo methods to avoid sampling uncertainty. Both experimental results based on the random sampling method and chronological sampling method are performed. Independent buoy observations are used to validate the calibration model. The results show that although there are obvious differences between the CFOSAT wavelength data and the field observations, the parameters observed by the satellite itself can effectively calibrate the data. In addition to the wavelength, nadir significant wave height, nadir wind speed, and the distance between the calibration point and satellite observation point are the most important parameters for the calibration. Accurate data from other sources, such as ERA5, would be helpful to further improve the calibration results. The variable contributing the most to the calibration effect is the mean wave period, which virtually provides relatively accurate wavelength information for the calibration network. These results verify the possibility of synchronous self-calibration for the CFOSAT wavelength data and provide a reference for the further calibration of the satellite products in other regions.

Integrative measurement analysis via machine learning descriptor selection for investigating physical properties of biopolymers in hairs

Integrative measurement analysis of complex subjects, such as polymers is a major challenge to obtain comprehensive understanding of the properties. In this study, we describe analytical strategies to extract and selectively associate compositional information measured by multiple analytical techniques, aiming to reveal their relationships with physical properties of biopolymers derived from hair. Hair samples were analyzed by multiple techniques, including solid-state nuclear magnetic resonance (NMR), time-domain NMR, Fourier transform infrared spectroscopy, and thermogravimetric and differential thermal analysis. The measured data were processed by different processing techniques, such as spectral differentiation and deconvolution, and then converted into a variety of “measurement descriptors” with different compositional information. The descriptors were associated with the mechanical properties of hair by constructing prediction models using machine learning algorithms. Herein, the stepwise model refinement via selection of adopted descriptors based on importance evaluation identified the most contributive descriptors, which provided an integrative interpretation about the compositional factors, such as α-helix keratins in cortex; and bounded water and thermal resistant components in cuticle. These results demonstrated the efficacy of the present strategy to generate and select descriptors from manifold measured data for investigating the nature of sophisticated subjects, such as hair.

Nutritional Metabolomics and the Classification of Dietary Biomarker Candidates: A Critical Review

Recent advances in metabolomics allow for more objective assessment of contemporary food exposures, which have been proposed as an alternative or complement to self-reporting of food intake. However, the quality of evidence supporting the utility of dietary biomarkers as valid measures of habitual intake of foods or complex dietary patterns in diverse populations has not been systematically evaluated. We reviewed nutritional metabolomics studies reporting metabolites associated with specific foods or food groups; evaluated the interstudy repeatability of dietary biomarker candidates; and reported study design, metabolomic approach, analytical technique(s), and type of biofluid analyzed. A comprehensive literature search of 5 databases (PubMed, EMBASE, Web of Science, BIOSIS, and CINAHL) was conducted from inception through December 2020. This review included 244 studies, 169 (69%) of which were interventional studies (9 of these were replicated in free-living participants) and 151 (62%) of which measured the metabolomic profile of serum and/or plasma. Food-based metabolites identified in ≥1 study and/or biofluid were associated with 11 food-specific categories or dietary patterns: 1) fruits; 2) vegetables; 3) high-fiber foods (grain-rich); 4) meats; 5) seafood; 6) pulses, legumes, and nuts; 7) alcohol; 8) caffeinated beverages, teas, and cocoas; 9) dairy and soya; 10) sweet and sugary foods; and 11) complex dietary patterns and other foods. We conclude that 69 metabolites represent good candidate biomarkers of food intake. Quantitative measurement of these metabolites will advance our understanding of the relation between diet and chronic disease risk and support evidence-based dietary guidelines for global health.

The exposome paradigm to predict environmental health in terms of systemic homeostasis and resource balance based on NMR data science

The environment, from microbial ecosystems to recycled resources, fluctuates dynamically due to many physical, chemical and biological factors, the profile of which reflects changes in overall state, such as environmental illness caused by a collapse of homeostasis. To evaluate and predict environmental health in terms of systemic homeostasis and resource balance, a comprehensive understanding of these factors requires an approach based on the "exposome paradigm", namely the totality of exposure to all substances. Furthermore, in considering sustainable development to meet global population growth, it is important to gain an understanding of both the circulation of biological resources and waste recycling in human society. From this perspective, natural environment, agriculture, aquaculture, wastewater treatment in industry, biomass degradation and biodegradable materials design are at the forefront of current research. In this respect, nuclear magnetic resonance (NMR) offers tremendous advantages in the analysis of samples of molecular complexity, such as crude bio-extracts, intact cells and tissues, fibres, foods, feeds, fertilizers and environmental samples. Here we outline examples to promote an understanding of recent applications of solution-state, solid-state, time-domain NMR and magnetic resonance imaging (MRI) to the complex evaluation of organisms, materials and the environment. We also describe useful databases and informatics tools, as well as machine learning techniques for NMR analysis, demonstrating that NMR data science can be used to evaluate the exposome in both the natural environment and human society towards a sustainable future.

Applicability of machine learning techniques in food intake assessment: A systematic review

The evaluation of food intake is important in scientific research and clinical practice to understand the relationship between diet and health conditions of an individual or a population. Large volumes of data are generated daily in the health sector. In this sense, Artificial Intelligence (AI) tools have been increasingly used, for example, the application of Machine Learning (ML) algorithms to extract useful information, find patterns, and predict diseases. This systematic review aimed to identify studies that used ML algorithms to assess food intake in different populations. A literature search was conducted using five electronic databases, and 36 studies met all criteria and were included. According to the results, there has been a growing interest in the use of ML algorithms in the area of nutrition in recent years. Also, supervised learning algorithms were the most used, and the most widely used method of nutritional assessment was the food frequency questionnaire. We observed a trend in using the data analysis programs, such as R and WEKA. The use of ML in nutrition is recent and challenging. Therefore, it is encouraged that more studies are carried out relating these themes for the development of food reeducation programs and public policies.

Relaxometric learning: a pattern recognition method for T2 relaxation curves based on machine learning supported by an analytical framework

Nuclear magnetic resonance (NMR)-based relaxometry is widely used in various fields of research because of its advantages such as simple sample preparation, easy handling, and relatively low cost compared with metabolomics approaches. However, there have been no reports on the application of the T₂ relaxation curves in metabolomics studies involving the evaluation of metabolic mixtures, such as geographical origin determination and feature extraction by pattern recognition and data mining. In this study, we describe a data mining method for relaxometric data (i.e., relaxometric learning). This method is based on a machine learning algorithm supported by the analytical framework optimized for the relaxation curve analyses. In the analytical framework, we incorporated a variable optimization approach and bootstrap resampling-based matrixing to enhance the classification performance and balance the sample size between groups, respectively. The relaxometric learning enabled the extraction of features related to the physical properties of fish muscle and the determination of the geographical origin of the fish by improving the classification performance. Our results suggest that relaxometric learning is a powerful and versatile alternative to conventional metabolomics approaches for evaluating fleshiness of chemical mixtures in food and for other biological and chemical research requiring a nondestructive, cost-effective, and time-saving method.

Correlation and association analyses in microbiome study integrating multiomics in health and disease

Correlation and association analyses are one of the most widely used statistical methods in research fields, including microbiome and integrative multiomics studies. Correlation and association have two implications: dependence and co-occurrence. Microbiome data are structured as phylogenetic tree and have several unique characteristics, including high dimensionality, compositionality, sparsity with excess zeros, and heterogeneity. These unique characteristics cause several statistical issues when analyzing microbiome data and integrating multiomics data, such as large p and small n, dependency, overdispersion, and zero-inflation. In microbiome research, on the one hand, classic correlation and association methods are still applied in real studies and used for the development of new methods; on the other hand, new methods have been developed to target statistical issues arising from unique characteristics of microbiome data. Here, we first provide a comprehensive view of classic and newly developed univariate correlation and association-based methods. We discuss the appropriateness and limitations of using classic methods and demonstrate how the newly developed methods mitigate the issues of microbiome data. Second, we emphasize that concepts of correlation and association analyses have been shifted by introducing network analysis, microbe-metabolite interactions, functional analysis, etc. Third, we introduce multivariate correlation and association-based methods, which are organized by the categories of exploratory, interpretive, and discriminatory analyses and classification methods. Fourth, we focus on the hypothesis testing of univariate and multivariate regression-based association methods, including alpha and beta diversities-based, count-based, and relative abundance (or compositional)-based association analyses. We demonstrate the characteristics and limitations of each approaches. Fifth, we introduce two specific microbiome-based methods: phylogenetic tree-based association analysis and testing for survival outcomes. Sixth, we provide an overall view of longitudinal methods in analysis of microbiome and omics data, which cover standard, static, regression-based time series methods, principal trend analysis, and newly developed univariate overdispersed and zero-inflated as well as multivariate distance/kernel-based longitudinal models. Finally, we comment on current association analysis and future direction of association analysis in microbiome and multiomics studies.

Comparative metabolomics implicates threitol as a fungal signal supporting colonization of Armillaria luteobubalina on eucalypt roots

Armillaria root rot is a fungal disease that affects a wide range of trees and crops around the world. Despite being a widespread disease, little is known about the plant molecular responses towards the pathogenic fungi at the early phase of their interaction. With recent research highlighting the vital roles of metabolites in plant root-microbe interactions, we sought to explore the presymbiotic metabolite responses of Eucalyptus grandis seedlings towards Armillaria luteobuablina, a necrotrophic pathogen native to Australia. Using a metabolite profiling approach, we have identified threitol as one of the key metabolite responses in E. grandis root tips specific to A. luteobubalina that were not induced by three other species of soil-borne microbes of different lifestyle strategies (a mutualist, a commensalist, and a hemi-biotrophic pathogen). Using isotope labelling, threitol detected in the Armillaria-treated root tips was found to be largely derived from the fungal pathogen. Exogenous application of d-threitol promoted microbial colonization of E. grandis and triggered hormonal responses in root cells. Together, our results support a role of threitol as an important metabolite signal during eucalypt-Armillaria interaction prior to infection thus advancing our mechanistic understanding on the earliest stage of Armillaria disease development. Comparative metabolomics of eucalypt roots interacting with a range of fungal lifestyles identified threitol enrichment as a specific characteristic of Armillaria pathogenesis. Our findings suggest that threitol acts as one of the earliest fungal signals promoting Armillaria colonization of roots.

InterSpin: Integrated Supportive Webtools for Low- and High-Field NMR Analyses Toward Molecular Complexity

InterSpin (http://dmar.riken.jp/interspin/) comprises integrated, supportive, and freely accessible preprocessing webtools and a database to advance signal assignment in low- and high-field NMR analyses of molecular complexities ranging from small molecules to macromolecules for food, material, and environmental applications. To support handling of the broad spectra obtained from solid-state NMR or low-field benchtop NMR, we have developed and evaluated two preprocessing tools: sensitivity improvement with spectral integration, which enhances the signal-to-noise ratio by spectral integration, and peaks separation, which separates overlapping peaks by several algorithms, such as non-negative sparse coding. In addition, the InterSpin Laboratory Information Management System (SpinLIMS) database stores numerous standard spectra ranging from small molecules to macromolecules in solid and solution states (dissolved in polar/nonpolar solvents), and can be searched under various conditions using the following molecular assignment tools. SpinMacro supports easy assignment of macromolecules in natural mixtures via solid-state ¹³C peaks and dimethyl sulfoxide-dissolved ¹H-¹³C correlation peaks. InterAnalysis improves the accuracy of molecular assignment by integrated analysis of ¹H-¹³C correlation peaks and ¹H-J correlation peaks of small molecules dissolved in D₂O or deuterated methanol, which supports easy narrowing down of metabolite candidates. Finally, by enabling database interoperability, SpinLIMS's client software will ultimately support scientific discovery by facilitating sharing and reusing of NMR data.