New Algorithm Combines Genes and Health Data to Profile Older Adults

Researchers in Spain have built and tested a multi-domain scoring algorithm designed to move nutrition advice for older adults beyond one-size-fits-all guidelines. The tool, described in a study published in the journal Genes & Nutrition, combines genetic risk scores, blood biomarkers, body composition, diet records, physical performance measures and digestive function into a single framework that assigns each person a set of health domain scores. When the researchers applied it to baseline data from 120 community-dwelling retirees enrolled in the MyFOOD4Senior study, the algorithm sorted the apparently healthy volunteers into three distinct health phenotypes: a reference or healthy group, a frailty-like group and a cardiovascular risk-like group.

The work, led by Cristina Álvarez-Martín and Elena Alonso-Aperte of CEU San Pablo University in Madrid, is explicitly a proof of concept rather than a validated clinical instrument. The authors stress that the algorithm was applied within a single cohort, lacks external validation and does not yet assess predictive performance or clinical effectiveness. Its purpose at this stage is to demonstrate that heterogeneous biological and lifestyle information can be translated into standardized, interpretable health scores that could later inform personalized dietary and physical activity recommendations for active older adults, pending longitudinal validation.

To construct the framework, the team first carried out a literature review to identify age-related conditions of interest, which anchored the definition of six health domains: Glucose Regulation, Nutritional Sufficiency, Cardiovascular Health, Physical Status, Digestive Health and Circadian Rhythms. Each domain was then operationalized through sub-domains integrating dietary intake, biochemical parameters, anthropometry, genetic information, body composition, physical activity, sleep patterns and gastrointestinal function. For every variable, evidence-based thresholds drawn from guidelines issued by the World Health Organization, the European Food Safety Authority, ESPEN and Spanish national bodies were used to classify individuals into optimal or suboptimal ranges.

A central component of the tool is a nutrigenetic test comprising 26 single nucleotide polymorphisms, or SNPs. From these, the researchers derived four candidate-SNP genetic risk scores covering type 2 diabetes, vitamin D deficiency, cardiovascular disease and strength deficiency. Genotypes were harmonized according to the number of protective alleles carried, and odds ratios from the literature were collapsed into three categories weighted 1.0, 0.5 and 0.25 to reflect high, intermediate and low effect sizes. Raw scores were standardized to a 0-to-10 scale, yielding high, intermediate and low genetic risk categories. The authors are careful to note that, given the categorical weighting approach and the absence of calibration in this population, these scores should be interpreted as exploratory genetic indices of susceptibility rather than validated measures of genetic risk.

The domain architecture assigns discrete points to each variable and sums them into an Overall Score with a maximum of 124 points. The Nutritional Sufficiency domain carries the largest weight at 48 points, combining dietary intake of energy, protein, vitamin D, iron, calcium and folates with biochemical measures including serum 25-hydroxyvitamin D, ferritin, transferrin saturation, haemoglobin, vitamin B12 and serum folate, plus the Mini Nutritional Assessment short form and genetic markers. The Cardiovascular Health domain reaches 26 points, integrating nine dietary variables, waist-to-hip ratio, body fat percentage, the cardiovascular disease genetic risk score and a composite risk estimate based on the 2021 European Society of Cardiology SCORE2 and SCORE2-OP charts. Physical Status spans 28 points and follows the EWGSOP2 sarcopenia framework, sequencing handgrip strength, chair stand performance, appendicular muscle mass and gait speed. Digestive Health contributes 6 points through lactase persistence, gluten intolerance susceptibility and diamine oxidase deficiency markers, while the Circadian Rhythms domain, built from chronotype questionnaires and CLOCK and PER3 variants, remains descriptive and is excluded from the Overall Score.

The study population consisted of 120 volunteers, mean age 69.9 years and 72.5 percent women, all retired, regular smartphone users and free of severe or limiting disease. At baseline the cohort showed an average body mass index of 29.9 kilograms per square metre, elevated systolic blood pressure of 143.3 millimetres of mercury, preserved muscle mass and high self-reported physical activity of roughly 2,507 MET-minutes per week. Despite an apparently favourable lifestyle profile, biochemical analysis revealed generalized vitamin D deficiency, with mean serum 25-hydroxyvitamin D of 23.4 nanomoles per litre, and suboptimal folate levels. Medication use was also notable, with 77.5 percent of participants taking cardiovascular drugs and 32.5 percent taking analgesics, while supplement use remained low.

Applying the algorithm produced a mean Overall Score of 79.86 points, about 64.4 percent of the maximum and placing the population in the moderate category. Glucose Regulation showed the highest relative adherence at 73.7 percent of its potential, followed by Digestive Health at 70.8 percent and Nutritional Sufficiency at 66.2 percent, while Physical Status at 61.9 percent and Cardiovascular Health at 56.5 percent scored lowest. A hierarchical cluster analysis using Ward’s method then identified three phenotypes with acceptable internal stability, reflected in an average silhouette width of 0.30, which the authors describe as supporting only an exploratory interpretation of the profiles.

The three clusters differed meaningfully across most domains. The reference or healthy phenotype, comprising 40 participants, achieved the highest scores everywhere, particularly in Glucose Regulation and Nutritional Sufficiency. The frailty-like phenotype, with 36 members, showed the lowest scores in Physical Status, Glucose Regulation and Nutritional Sufficiency, differences that were significant against both other clusters. The cardiovascular risk-like phenotype, the largest group at 44 participants, maintained high Nutritional Sufficiency comparable to the healthy group yet recorded the lowest Cardiovascular Health score. Notably, Digestive Health did not differentiate the clusters. The frailty-like group also had the lowest Healthy Eating Index at 66.8, significantly below the other two clusters, whereas the cardiovascular risk-like group maintained a high dietary quality, illustrating that certain vulnerabilities can persist even when dietary patterns appear adequate.

Secondary, exploratory comparisons added context to the phenotypes. Educational level differed significantly across clusters, with 55 percent of the healthy group having medium or high education, while 75 percent of participants in each of the other two clusters had low educational background. Antidepressant use was reported by 31.8 percent of the cardiovascular risk-like cluster versus 10.0 percent of the reference cluster, and cardiovascular medication use reached 86.1 percent in the frailty-like cluster compared with 62.5 percent in the reference group. Supplement use patterns also varied, with cardiovascular supplements used only in the healthy cluster and digestive health supplements most common in the frailty-like group, though the authors treat all medication and supplement comparisons as strictly exploratory.

The authors position the framework alongside prior personalized nutrition research, including the Food4Me study, which found personalized advice based on phenotypic and genetic data more effective than conventional guidelines, and work showing that metabolic responses to identical nutrients vary substantially between individuals. They argue that multidimensional scoring can surface risks that single markers or dietary assessment alone would miss, such as a person with adequate nutritional sufficiency who still falls into the cardiovascular risk-like phenotype. However, they catalogue substantial limitations: the selected cohort of healthy, smartphone-using retirees limits generalizability; the rule-based weighting strategy has not been calibrated; the discrete genetic weighting could underestimate SNPs lacking meta-analyses; the cardiovascular component omitted disease-specific pathways for diabetes and chronic kidney disease; and no prospective intervention data exist yet. Longitudinal follow-up from the full one-year trial, registered as NCT07461883, will be needed to determine whether recommendations derived from these scores affect chronic disease development and quality of life in older adults.

Subject of Research: A multi-domain precision nutrition algorithm integrating genetic risk scores and health phenotyping in older adults

Article Title: A multi-domain algorithm towards precision nutrition: integrating genetic risk scores and phenotypic diversity in the elderly, the MyFOOD4Senior study

Article References: Álvarez-Martín, C., Benavent, M., Úbeda, N., García-González, Á., Martínez-Saez, N., González, M. P., Achón, M., Fajardo, V., de la Iglesia, R., & Alonso-Aperte, E. (2026). A multi-domain algorithm towards precision nutrition: integrating genetic risk scores and phenotypic diversity in the elderly, the MyFOOD4Senior study. Genes & Nutrition, 21(1), Article 23. https://doi.org/10.1186/s12263-026-00822-y

Image Credits: AI Generated

DOI: 10.1186/s12263-026-00822-y

Keywords: precision nutrition, older adults, genetic risk scores, nutrigenetics, health phenotypes, MyFOOD4Senior, sarcopenia, cardiovascular risk, vitamin D deficiency, cluster analysis, personalized nutrition, healthy ageing