The European project EprObes employs clinical and epigenetic data to tailor risk predictions, enabling early interventions.
Obesity has emerged as a significant public health threat globally, with its prevalence rising over recent decades and associated comorbidities profoundly affecting individual well-being and reducing life expectancy. Despite research advancements, current obesity treatments have shown limited results, highlighting the urgent need for effective preventive strategies to mitigate metabolic complications linked to overweight throughout life.
In this context, various teams from the Biomedical Research Networking Center (CIBER) at the University of Granada (UGR) have developed an artificial intelligence (AI) model to predict metabolic disorder risks in obese children. The work, published in the journal Artificial Intelligence in Medicine, stands out for integrating clinical and epigenetic data to estimate the risk of metabolic complications in the coming years.
This study has identified that children with metabolic disorders during puberty exhibit distinctive clinical and epigenetic patterns from the prepubertal stage. Researchers emphasize that implementing this AI model in hospitals could facilitate early detection of metabolic risks, allowing timely interventions through pharmacological treatments or lifestyle changes, thereby preventing metabolic diseases. Additionally, this strategy could reduce comorbidities associated with obesity and has the potential to lower costs for the public health system, according to the study's authors.
The study has been coordinated by personnel from the Obesity and Nutrition Pathophysiology area of CIBER (CIBEROBN) at the University of Granada, the Biosanitary Research Institute (ibs.GRANADA), and the Andalusian Interuniversity Institute in Data Science and Computational Intelligence (DaSCI), among other institutions. It has also been developed with financial support from the Carlos III Health Institute and the European project EprObes (Preventing lifetime obesity by early risk-factor identification, prognosis, and intervention). This European program primarily aims to prevent obesity by early identification of risk factors, providing accurate prognosis, and facilitating timely interventions.
Obesity development in adults is closely related to early maturation events, including physiological and psychological factors occurring during gestational, childhood, and adolescent stages; however, these determinants remain poorly understood. Identifying early pathogenic mechanisms and metabolic disease markers is essential for designing active prevention strategies and personalized plans for managing body weight in later life stages.
A relevant and still insufficiently explored aspect is how pathogenic mechanisms and susceptibility to obesity vary by gender. This lack of knowledge could limit the effectiveness of preventive measures and treatments designed to address both obesity and its metabolic complications.
In this regard, the EprObes project is a multidisciplinary, patient-centered initiative that combines clinical studies at various developmental stages with research in mental health, behavior, lifestyle, and cognition. It also incorporates mechanistic analyses using advanced preclinical models to establish effective strategies for actively preventing obesity throughout life, with a special focus on critical development periods, from the prenatal stage (including the periconceptional period) to puberty, as well as factors influencing eating behaviors.
Through multi-omic studies and comprehensive data analysis supported by bioinformatics technologies and artificial intelligence, EprObes aims to design personalized preventive measures and promote lifestyle interventions at key developmental moments. These actions target preventing excess body weight and metabolic complications throughout life, with specific attention to both sexes.
The AI model combines traditional data, such as body mass index and hormone levels of leptin and adiponectin, with new genetic markers in key genes like HDAC4, PTPRN2, MATN2, RASGRF1, and EBF1. One of the most innovative features of this model is its design as an explainable AI, allowing healthcare professionals to interpret its functioning and understand the basis of its predictions, thereby facilitating its integration into clinical practice.
This combination of data enables not only precise risk prediction but also a greater understanding of how the model processes variables, allowing for more effective application in clinical settings,” stated Álvaro Torres, a researcher at CIBEROBN.