Descriptive characteristics of the NUTRiMDEA population
The phenotypic characteristics of participants in the NUTRiMDEA study were categorized by gender (female and male) and age (<40 years and ≥40 years), and the data are shown in Tables 1, 2, and 3. Significant differences were found in age, educational background, family situation, number of meals per day and snacking habits, napping habits, weight, BMI, and PCS12. A higher proportion of female and younger participants reported having a university education, but younger participants reported higher student rates and lower employment rates. Young men were more likely to live alone, whereas older men reported living with a spouse or children. They reported that older men had a higher prevalence of cardiometabolic diseases. Regarding family history, women reported a higher prevalence of familial hypertension (HBP) and dyslipidemia. Younger women reported more depression, while older men reported lower rates. Considering lifestyle, young men reported higher rates of smoking habits, while young women tended to have lower rates of smoking habits, but more frequent meals and snacking habits. Regardless of gender, younger people reported replacing food with snacks more often and drinking more water than older participants. Subjects under 40 years of age reported longer naps (60 minutes or more) and longer sleep hours at night compared to subjects over 40 years of age. The data revealed significant differences in physical activity by age and gender. Older people say they spend less time sitting, but older women spend more time sitting than older men. Regardless of age, men self-reported being more moderately active than women. Older men reported participating in more light physical activity, while younger men reported participating in more vigorous physical activity. Older women engaged in less strenuous physical activity. Men obtained higher levels of total METs-min/week than women, regardless of age. Older men had higher BMI, and younger women had lower BMI. Additionally, women and older adults had higher scores on the MDS14. Regarding HRQoL, regardless of gender, younger people had higher PCS12 scores, men and older people had higher MCS12 scores, and younger women had lower MCS12 scores.
Table 1 Sociodemographic and health characteristics of NUTRiMDEA participants distributed by gender and age, n (%). Table 2 Lifestyle characteristics of NUTRiMDEA participants distributed by gender and age, n (%). Table 3. Nutrition, physical activity and HRQoL characteristics of NUTRiMDEA participants, distribution and mean (SD) by gender and age.
Exploratory factor analysis and variable selection
After exploratory factor analysis and scree plot testing of the 62 variables, we obtained 19 factors with eigenvalues greater than 1, explaining 57.5% of the total variance. The resulting KMO measurement is 0.7578, which is above the KMO rule of thumb cutoff. This value should typically be greater than 0.60 for sampling adequacy. Subsequently, an orthogonal rotation procedure (promax rotation), visual inspection of the correlation matrix was performed, and absolute factor loadings ≥0.30 were considered significant for each factor (Supplementary Figure S2). Variables related to HRQoL include Factors (1) which determine age, gender, cardiometabolic health, some aspects of HRQoL, and smoking habits; (2) Factor 3, which determines the Mediterranean diet, smoking habits, and It consisted of questions related to aspects of. Rate of HRQoL was included in factor 4. Factors 5 and 6 were dominated by family medical history and physical activity. Living situation was associated with factors 5 and 7, and obesity and smoking prevalence were significant for factor 8. Factor 9 consisted of sleep and smoking habits. Factor 10 included the food aspect. Factor 11 was related to marital life, sleeping, and snacking habits. For factor 12, living alone, living with others, and occupation were the most involved variables. Ethnicity and use of olive oil as the main fat in the diet were significant for factor 13. Gender, diabetes prevalence, pulse intake, and physical exercise were included in factor 14. Factor 15 consisted of questions regarding napping habits and feeling calm and quiet. in HRQoL questionnaire. Dyslipidemia and salt added to the dish significantly contributed to factor 16. Living with older adults and the prevalence of diabetes were associated with factor 17. On the other hand, living with an elderly person and consumption of white meat rather than red or processed meat were more important for factor 18. 19 were associated with education, wine consumption, weekly seafood intake, and consumption of more white meat than red or processed meat. Hierarchical cluster analysis resulted in five life metabotypes, which are illustrated descriptively in a dendrogram (Figure 2).
Figure 2
Dendrogram. Phenotypic description of cluster analysis. The dendrogram graphically represents the number of groups, as seen on the horizontal axis. Within each group, the corresponding population number (n) is displayed. Colors indicate the five best clusters used to characterize the phenotype.
Clustering information
The clustering method yielded five clusters. After analyzing the variables of each cluster, phenotypic characteristics were identified (Table 4 and Supplementary Tables S1 and S2). Cluster 1, labeled “Westernized Millennials,” included 967 participants, primarily between the ages of 18 and 40, who were mostly female, white, college-educated, and employed or I was a student. The majority reported living with other people. This cluster did not show a significant prevalence of cardiometabolic disease or family history, but almost half reported frequent sadness and depression. Respondents in this cluster declared the highest consumption of red/processed meat, moderate to low adherence to the Mediterranean diet, and moderate physical activity level. This cluster also had the highest proportion of non-smokers, while smokers reported smoking the least amount of cigarettes per day. PCS12 scores were moderate, whereas MCS12 scores were moderate to low.
Cluster 2 “Health” included 10,616 volunteers. This group is reported to be primarily middle-aged (40-55 years), with a slightly higher proportion of women than men, and a predominance of whites over Hispanics. Most participants were university-educated, living with their spouse and children, were employed, maintained a normal weight, had no cardiometabolic disease, but had family members with HBP and dyslipidemia. He had a history of having no snacking habits, was a non-smoker, and declared no symptoms of depression. This population obtained high adherence to a Mediterranean diet and high levels of moderate physical activity, achieving moderate scores on HRQoL.
Cluster 3, named “Mediterranean Adolescents and Adults”, consisted of 2013 participants. This group consisted primarily of young and middle-aged adults (25–55 years), with higher proportions of women and whites. The participants were people who had graduated from university, were employed, and were living alone. This cluster was found to have the lowest prevalence of obesity, diabetes, and family history compared to other clusters. They self-reported minimal symptoms of depression and best adherence to a Mediterranean diet. Their physical activity level was moderate, but they spent a lot of time sedentary. Most were non-smokers and had moderate HRQoL scores.
Cluster 4, “premorbid” (n = 600), includes a diverse age group, is predominantly female, is comprised of both whites and Hispanics, and many are college educated. Participants described living with an elderly person, being employed, a student, or unemployed. Many were overweight and reported a family history of HBP or dyslipidemia. Approximately half had experienced depression and the majority were non-smokers. They self-reported frequently snacking and adding salt. Adherence to the Mediterranean diet was lower, but physical activity was slightly higher. PCS12 was moderate, while MCS12 was relatively low.
Cluster 5 “premorbid” (n = 312) included middle-aged to older adults, more women, and Caucasians. Most are declared to have a university or professional education and live with their partner and children. This cluster included varying proportions of employed, retired, unemployed, and disabled people. Participants had a balanced distribution between normal weight and overweight, with a significant prevalence of obesity. This cluster showed the highest prevalence of cardiometabolic diseases and family history, as well as more depressive symptoms. They self-reported sleeping less than 7 to 8 hours per night and had the lowest water intake. Sedentary time ranged from 5 to 7 hours per day to 8 to 10 hours per day. Adherence to the Mediterranean diet, physical activity, and PCS12 were low, whereas MCS12 was moderate.
Table 4 Description of the most relevant characteristics of participants based on the variables most important to the computational phenotyping algorithm.
Computational algorithm development
After performing forward stepwise regression, the following variables were identified in the final model: age (18-25 years / 25-40 years / 40-55 years / 55-70 years / >70 years), gender (female / male / gender not specified), T-shirt size (XS / S / M / L / XL / XXL), occupation (unemployed / student / disabled / retired / housewife / employed), ethnicity (Caucasian/European/Hispanic/Latinx/African/Asian)/Mestizo/Other/prefer not to specify), live alone, live with someone older than you, or live with others (yes/no), sleep duration on weekdays (<5 hours per day / 5-6 hours per day / 7-8 hours per day / 9-10 hours per day / >10 hours per day) , Obesity and Diabetes Prevalence (Yes/No), Familial Obesity, Diabetes and HBP (Yes/No/DKDA), Water (1-2 glasses/3-4 glasses per day) per day/5 per day ~6 cups / 7-8 cups per day / 9-10 cups per day / 10 or more cups per day), number of meals (1 or 2 meals per day / 3 meals per day / 4 meals per day) 1 day / 1 5 meals per day / 6 or more meals per day), red and processed meats, butter/cream/margarine (never or rarely / 1 serving per day / 2 or more servings per day), sugar-sweetened beverages (never or rarely (none / 1-2 servings per day / 3 or more servings per day), seafood (never or rarely / 1-2 servings per week / 3 or more servings per week), preferring white meat to red meat (yes / no), moderate physical activity (hours/week), self-perception of health (very good / very good / good / fair / poor), moderate activity and limited ability to go up and down stairs (yes, very limited / yes, a little limited / no, not limited) all), underachievement due to physical health or emotional problems (yes / no), work or Other activities were limited (yes/no), emotional issues prevented careful work (yes/no), pain (not) all/slightly/moderately/quite/extremely ), depressed and depressed, and calm and peaceful (always / most of the time / a lot of the time / sometimes / a little bit of the time / no time).
The beta coefficients (β) of the variables selected for the development of the computational algorithm and the model variance contribution (R2) of each variable are shown in Supplementary Table S3. The variables that contributed the most to the model were living with an older person (27.8%), living alone (12.9%), living with others (6.6%), and having moderate activity limitations (1.3%).
A calculation algorithm was obtained through a calculation formula that can estimate the classification of each life metabotype (Figure 3). When we calculated the probability that participants would be classified into different groups or clusters using a random forest algorithm, we obtained the following results. Cluster 1 was classified into that group 81.9% of the time, Cluster 2 was 94.6%, and Cluster 3 was 82.5%. %, cluster 4 is 79.7%, cluster 5 is 77.5% (Supplementary Table S4).
Figure 3
Computational algorithms for classifying phenotypes. Intercept = 3.5092.
