A novel carbon isotope biomarker for dietary sugar

Machine learning modeling practices to support the principles of AI and ethics in nutrition research

BACKGROUND

Nutrition research is relying more on artificial intelligence and machine learning models to understand, diagnose, predict, and explain data. While artificial intelligence and machine learning models provide powerful modeling tools, failure to use careful and well-thought-out modeling processes can lead to misleading conclusions and concerns surrounding ethics and bias.

METHODS

Based on our experience as reviewers and journal editors in nutrition and obesity, we identified the most frequently omitted best practices from statistical modeling and how these same practices extend to machine learning models. We next addressed areas required for implementation of machine learning that are not included in commercial software packages.

RESULTS

Here, we provide a tutorial on best artificial intelligence and machine learning modeling practices that can reduce potential ethical problems with a checklist and guiding principles to aid nutrition researchers in developing, evaluating, and implementing artificial intelligence and machine learning models in nutrition research.

CONCLUSION

The quality of AI/ML modeling in nutrition research requires iterative and tailored processes to mitigate against potential ethical problems or to predict conclusions that are free of bias.

A Systematic Review of Metabolomic Biomarkers for the Intake of Sugar-Sweetened and Low-Calorie Sweetened Beverages

Intake of added sugars (AS) is challenging to assess compared with total dietary sugar because of the lack of reliable assessment methods. The reliance on self-reported dietary data in observational studies is often cited as biased, with evidence of AS intake in relation to health outcomes rated as low to moderate quality. Sugar-sweetened beverages (SSBs) are a major source of AS. A regular and high intake of SSBs is associated with an overall poor diet, weight gain, and cardiometabolic risks. An elevated intake of low-calorie sweetened beverages (LCSBs), often regarded as healthier alternatives to SSBs, is also increasingly associated with increased risk for metabolic dysfunction. In this review, we systematically collate evidence and provide perspectives on the use of metabolomics for the discovery of candidate biomarkers associated with the intake of SSBs and LCSBs. We searched the Medline, Embase, Scopus, and Web of Science databases until the end of December 2020. Seventeen articles fulfilled our inclusion criteria. We evaluated specificity and validity of the identified biomarkers following Guidelines for Biomarker of Food Intake Reviews (BFIRev). We report that the ¹³C:¹²C carbon isotope ratio (δ¹³C), particularly, the δ¹³C of alanine is the most robust, sensitive, and specific biomarker of SSBs intake. Acesulfame-K, saccharin, sucralose, cyclamate, and steviol glucuronide showed moderate validity for predicting the short-term intake of LCSBs. More evidence is required to evaluate the validity of other panels of metabolites associated with the intake of SSBs.

Biomarkers of intake for coffee, tea, and sweetened beverages

Non-alcoholic beverages are important sources of nutrients and bioactive compounds that may influence human health and increase or decrease the risk of chronic diseases. A wide variety of beverage constituents are absorbed in the gut, found in the systemic circulation and excreted in urine. They may be used as compliance markers in intervention studies or as biomarkers of intake to improve measurements of beverage consumption in cohort studies and reveal new associations with disease outcomes that may have been overlooked when using dietary questionnaires. Here, biomarkers of intake of some major non-alcoholic beverages-coffee, tea, sugar-sweetened beverages, and low-calorie-sweetened beverages-are reviewed. Results from dietary intervention studies and observational studies are reviewed and analyzed, and respective strengths and weaknesses of the various identified biomarkers discussed. A variety of compounds derived from phenolic acids, alkaloids, and terpenes were shown to be associated with coffee intake and trigonelline and cyclo(isoleucylprolyl) showed a particularly high specificity for coffee intake. Epigallocatechin and 4'-O-methylepigallocatechin appear to be the most sensitive and specific biomarkers for green or black tea, while 4-O-methylgallic acid may be used to assess black tea consumption. Intake of sugar-sweetened beverages has been assessed through the measurement of carbon-13 enrichment of whole blood or of blood alanine in North America where sugar from sugarcane or corn is used as a main ingredient. The most useful biomarkers for low-calorie-sweetened beverages are the low-calorie sweeteners themselves. Further studies are needed to validate these biomarkers in larger and independent populations and to further evaluate their specificity, reproducibility over time, and fields of application.

Validation of biomarkers of food intake-critical assessment of candidate biomarkers

Biomarkers of food intake (BFIs) are a promising tool for limiting misclassification in nutrition research where more subjective dietary assessment instruments are used. They may also be used to assess compliance to dietary guidelines or to a dietary intervention. Biomarkers therefore hold promise for direct and objective measurement of food intake. However, the number of comprehensively validated biomarkers of food intake is limited to just a few. Many new candidate biomarkers emerge from metabolic profiling studies and from advances in food chemistry. Furthermore, candidate food intake biomarkers may also be identified based on extensive literature reviews such as described in the guidelines for Biomarker of Food Intake Reviews (BFIRev). To systematically and critically assess the validity of candidate biomarkers of food intake, it is necessary to outline and streamline an optimal and reproducible validation process. A consensus-based procedure was used to provide and evaluate a set of the most important criteria for systematic validation of BFIs. As a result, a validation procedure was developed including eight criteria, plausibility, dose-response, time-response, robustness, reliability, stability, analytical performance, and inter-laboratory reproducibility. The validation has a dual purpose: (1) to estimate the current level of validation of candidate biomarkers of food intake based on an objective and systematic approach and (2) to pinpoint which additional studies are needed to provide full validation of each candidate biomarker of food intake. This position paper on biomarker of food intake validation outlines the second step of the BFIRev procedure but may also be used as such for validation of new candidate biomarkers identified, e.g., in food metabolomic studies.