Related Experiment Video
Updated: Jun 15, 2025
Author Spotlight: Integrated Multi-Omics Analysis for Unveiling Multicellular Immune Signatures in Clinical Heart Attack Cohorts
Published on: September 20, 2024
Multiomics and eXplainable artificial intelligence for decision support in insulin resistance early diagnosis: A
Álvaro Torres-Martos1, Augusto Anguita-Ruiz2, Mireia Bustos-Aibar3
1Department of Biochemistry and Molecular Biology II, School of Pharmacy, "José Mataix Verdú" Institute of Nutrition and Food Technology (INYTA) and Center of Biomedical Research, University of Granada, Granada, 18071, Spain; Instituto de investigación Biosanitaria ibs.GRANADA, Granada, 18012, Spain; CIBER de Fisiopatología de la Obesidad y Nutrición (CIBEROBN), Instituto de Salud Carlos III, Madrid, 28029, Spain.
Early prediction of insulin resistance in pediatric obesity is crucial. An explainable AI system using multi-omics and clinical data accurately identified children at risk, highlighting key biomarkers for intervention.
Area of Science:
- Pediatric Endocrinology
- Computational Biology
- Artificial Intelligence in Medicine
Background:
- Pediatric obesity increases long-term cardiometabolic risk, with insulin resistance being a key link.
- Insulin resistance developing during or after puberty is harder to reverse, emphasizing the need for early detection.
- Predictive systems require longitudinal data and integration of diverse factors for robust inference.
Purpose of the Study:
- To develop an explainable AI decision support system for early diagnosis of insulin resistance in pre-pubertal children with obesity.
- To leverage multi-omics (genomics, epigenomics) and clinical data for accurate prediction.
- To provide interpretable insights into the factors driving the diagnostic decisions.
Main Methods:
- Utilized a longitudinal cohort of 90 children, integrating multi-omics and clinical data from the pre-pubertal stage.
- Employed a robust machine learning pipeline with various data pre-processing techniques and algorithms.
- Implemented SHapley Additive exPlanations (SHAP) for model interpretability at global and local levels.
Main Results:
- The AI system achieved high predictive performance with an AUC and G-mean of 0.92.
- Identified significant biomarkers for insulin resistance, including Body Mass Index z-score, leptin/adiponectin ratio, and novel gene methylation patterns (e.g., HDAC4, PTPRN2).
- SHAP analysis provided clear explanations for individual predictions, enhancing clinical trust and utility.
Conclusions:
- Integrating multi-omics data with explainable AI offers a powerful approach for early insulin resistance detection in pediatric obesity.
- The developed system demonstrates significant potential for clinical decision support in preventing long-term cardiometabolic complications.
- Novel biomarkers identified may pave the way for targeted interventions and improved patient outcomes.
