Food bioactive small molecule databases: Deep boosting for the study of food molecular behaviors

AI nutritionist: Intelligent software as the next generation pioneer of precision nutrition

With the rapid development of information technology and artificial intelligence (AI), people have acquired the abilities and are encouraged to develop intelligent tools and software, which begins to shed light on intelligent and precise food nutrition. Despite the rapid development of such software, disparities still exist in terms of methodology, contents, and implementation strategies. Hence, a set of panoramic profiles is urgently needed to elucidate their values and guide their future development. Here a comprehensive review was conducted aiming to summarize and compare the objects, contents, intelligent algorithms, and functions realized by the already released software in current research. Consequently, 177 AI nutritionists in recent years were collected and analyzed. The advantages, limitations, and trends concerning their application scenarios were analyzed. It was found that AI nutritionists have been gradually advancing the production modes and efficiency of food recognition, dietary recording/monitoring, nutritional assessment, and nutrient/recipe recommendation. Most AI nutritionists have a relatively low level of intelligence. However, new trends combining advanced AI algorithms, intelligent sensors and big data are coming with new applications in real-time and precision nutrition. AI models concerning molecular-level behaviors are becoming the new focus to drive AI nutritionists. Multi-center and multi-level studies have also gradually been realized to be necessary.

HS-SPME/GC×GC-TOFMS-Based Flavoromics and Antimicrobial Properties of the Aroma Components of Zanthoxylum motuoense

Zanthoxylum motuoense Huang, native to Tibet, China, is a newly discovered Chinese prickly ash, which, recently, has increasingly attracted the attention of researchers. In order to understand its volatile oil compositions and flavor characteristics, and to explore the flavor difference between Z. motuoense and the common Chinese prickly ash sold in the market, we analyzed the essential oils of Z. motuoense pericarp (MEO) using HS-SPME/GC×GC-TOFMS coupled with multivariate data and flavoromics analyses. The common commercial Chinese prickly ash in Asia, Zanthoxylum bungeanum (BEO), was used as a reference. A total of 212 aroma compounds from the 2 species were identified, among which alcohols, terpenoids, esters, aldehydes, and ketones were the major compounds. The predominant components detected from MEO were citronellal, (+)-citronellal, and β-phellandrene. Six components-citronellal, (E,Z)-3,6-nonadien-1-ol, allyl methallyl ether, isopulegol, 3,7-dimethyl-6-octen-1-ol acetate, and 3,7-dimethyl-(R)-6-octen-1-ol-could be used as the potential biomarkers of MEO. The flavoromics analysis showed that MEO and BEO were significantly different in aroma note types. Furthermore, the content differences of several numb taste components in two kinds of prickly ash were quantitatively analyzed using RP-HPLC. The antimicrobial activities of MEO and BEO against four bacterial strains and nine plant pathogenic fungi were determined in vitro. The results indicated that MEO had significantly higher inhibitory activities against most microbial strains than BEO. This study has revealed the fundamental data in respect of the volatile compound properties and antimicrobial activity of Z. motuoense, offering basic information on valuable natural sources that can be utilized in the condiment, perfume, and antimicrobial sectors.

Predictive and explanatory themes of NOAEL through a systematic comparison of different machine learning methods and descriptors

No observed adverse effect level (NOAEL) is an identified dose level which used as a point of departure to infer a safe exposure limit of chemicals, especially in food additives and cosmetics. Recently, in silico approaches have been employed as effective alternatives to determine the toxicity endpoints of chemicals instead of animal experiments. Several acceptable models have been reported, yet assessing the risk of repeated-dose toxicity remains inadequate. This study established robust machine learning predictive models for NOAEL at different exposure durations by constructing high-quality datasets and comparing different kinds of molecular representations and algorithms. The features of molecular structures affecting NOAEL were explored using advanced cheminformatics methods, and predictive models also communicated the NOAEL between different species and exposure durations. In addition, a NOAEL prediction tool for chemical risk assessment is provided (available at: https://github.com/ifyoungnet/NOAEL). We hope this study will help researchers easily screen and evaluate the subacute and sub-chronic toxicity of disparate compounds in the development of food additives in the future.