The Osaka Prefectural Government has formulated an inspection method for the four elements of frailty: nutrition, body function, oral cavity, and social activity decline. Since nutrition is primarily influenced by food intake, developing and utilizing a food ontology is essential for scientific research on frailty. Recent studies have also elucidated the relationship between food, gut bacteria, and disease. Foods are transformed into nutrients through metabolism, and in the process, there are changes in various genes and proteins, which are also associated with diseases in relation to gut bacteria. We aim to contribute to better prevention of frailty by constructing a knowledge graph consisting of these elements.
frailty and sarcopenia may afect not only the elderly but also the working-age population, early prevention is necessary. To prevent frailty and sarcopenia, it is important to improve lifestyle habits such as diet and exercise. People who perceive themselves as heavier than they actually are may have low muscle mass, which should be taken into account.
A knowledge graph with guidelines derived from previous frailty prevention studies is expected to contribute to more accurate frailty prevention. Genome-wide association studies (GWAS) have identified single nucleotide polymorphisms (SNPs) associated with frailty. The presence of SNPs suggests that individual diferences in frailty prevention may occur. It has also been reported that the gut microbiota environment, which is influenced by food intake, is involved in the expression of messenger RNAs (mRNAs) and microRNAs (miRNAs), and vice versa, the expression of mRNAs and miRNAs, which are influenced by food intake, afects the gut microbiota environment. This suggests that the possible knowledge graph for frailty prevention can be very complex with other factors.
Weakness is said to be caused by the decline of four functions: nutrition, body function, oral cavity, and social activity. In Osaka Prefecture, the assessment includes whether individuals can form a loop with the thumb and index finger of both hands, and whether they can stand up from a chair on one leg. Whether or not the patient has a complete meal with staple food, main dishes, and side dishes. Whether or not the patient swallows when eating or drinking tea or soup. Whether or not the patient goes out once a week. The checklist consists of the following ifve items. The knowledge graph is created according to this list.
The above mentioned indicators of frailty diagnosis include food, necessitating a Japanese
food ontology that cabe processed mechanically. FoodOn[
Associations between food, gut bacteria and disease were extracted from the databases listed below.
• Disease name and DOID from Disease ontology
(https://disease-ontology.org/) • Gut bacteria name and NCBI ID from NCBI taxonomy
(https://www.ncbi.nlm.nih.gov/taxonomy) • Gut bacteria disease interaction type from gutMDisorder database
(http://bio-annotation.cn/gutMDisorder/) • Gut bacteria edge type and weight from MIND database
(http://www.microbialnet.org/mind_home.html)
We are building a knowledge graph using the constructed food ontology and various other factors related to frailty, but since there is some cohort data dependence, we have not yet reached a concrete inference. The constructed knowledge graph schema is Figure 2. We proposed that individual diferences in genes that vary with food metabolism may influence frailty.
The four factors that traditionally define frailty are body function, social activity, oral cavity, and nutrients. A food component was added because nutrients are obtained through food metabolism, but food intake also causes variation in genes and gut bacteria, which afect frailty in terms of disease improvement. Individual diferences due to genetic variants must also be taken into account. In addition, the factor of sleep should also play a role in frailty, although cohort data may not be suficient. Therefore, the knowledge graph on frailty could be further complicated, as shown in Figure 2, and could contribute to precise frailty prevention eforts. It is dificult to predict frailty using the knowledge graph with only a database of known reports, and additional data is needed. These are still in the process of being built and the various processes listed in future work need to be implemented. As this system matures, it is expected to elucidate the various associations and mechanisms between food and disease.
We thank to Dr. Tatsuya Kushida (RIKEN), Assist. Prof. Chioko Nagao (IPR), Assoc. Prof. Hideki Hatanaka (DBCLS), Prof. Kouji Kozaki (OECU), a collaborator in the development of the FGNHNS, and the members of the Artificial Intelligence Center for Health and Biomedical Research (ArCHER).