Exploring the use of machine learning in metabolomics and personalised nutrition
Chronic diseases such as cardiovascular disease, type 2 diabetes, and cancers are among the largest public health burdens in modern society. Important for the prevention of these diseases are key modifiable risk factors, such as dietary intake, physical activity and body composition (Ng 2020). The transition between childhood and adulthood (ages 12-25) is a crucial, yet often overlooked, period for establishing long-term health behaviour patterns (Nelson et al. 2008). During this time, health risks are diverse and rapidly changing, in addition to biological growth and development, which are still occurring intensively (Patton 2016). Maintaining healthy diets is a challenge for adolescents (ages 10–19) and young adults (ages 18–25). According to national health surveys, the majority of Irish adolescents do not meet dietary recommendations (Rippin 2019). Dietary behaviours also tend to worsen during early adulthood, which is typically classified for young adults by a transition into independent living. Over time, these poor dietary behaviours can lead to an increased risk of chronic disease across the life course (Blichfeldt and Gram 2013). Thus, there is a need for the timely implementation of effective interventions to help improve dietary behaviours in young people across Ireland. Previous nutritional interventions targeting adolescents and young adults, often involving the administration of general dietary advice and education, have seen relatively little long-term success (Whatnall 2021). Precision Nutrition (PN) is an emerging field that concerns the use of personal information to generate nutritional advice that, hypothetically, leads to superior health outcomes than generic advice (Kirk 2022). This idea rests on the basis that inter-individual variability in genomic, epigenetic, microbiome, and environmental factors, ultimately calls for specific nutritional requirements that population-level guidelines cannot capture. Due to the complexity of the data that represent these factors, machine learning (ML) has been recognised as an indispensable tool for data generation and analysis in PN (NIH 2022). While an increasing number of studies have demonstrated the potential for PN to predict individual responses to dietary interventions, further experimental investigations are required to establish whether PN could be a more effective alternative to generic healthy eating recommendations. Metabolomics is an area closely related to PN. ML has potential to analyse metabolomics data for applications such as disease prediction and identification of biomarkers (Reel 2021). Clinical biomarkers to evaluate chronic disease risk may be of particular importance in adolescents and young adults, especially considering the preventive potential in these age groups. Despite its potential, the uptake of ML methods in nutrition research has been slow, in part due to difficulties with implementation, and the lack of sufficiently large datasets. However, the growing number of publicly available datasets in key areas such as PN and metabolomics has provided an avenue for the advancement of ML techniques and applications in this field. Thus, the aim of the present research is to investigate the uses of ML in metabolomics and personalised nutrition, with a particular focus on its ability to help improve dietary behaviours in young people.