Integrating Multi-Omics Data for Advanced Diabetes Prediction and Understanding

Diabetes is a major global health issue, requiring new approaches to understanding its complex etiology and improving predictions of effective management and treatment. Current methods typically use centralized models, which have privacy concerns and cannot fully capture correlations in multimodal datasets. To address these problems, we introduce a new model using integrated learning and causal theory. Our system uses three datasets: the Pima Indian Diabetes Dataset (PIDD), 130 U.S. Appl. Clinical Diabetes Dataset, and physiological summaries. For analysis in our proposed method, we use various machine learning methods with ensemble method. Here we achieve an accuracy of 93% for PIDD 92% for US 130 clinical datasets and 90% for physically collected datasets. By integrating genetic, proteomic, and metabolomic data, we better understand diabetic mechanisms and potential therapeutic targets. Our model addresses the limitations of current methods, improves prediction accuracy, and data privacy, and provides insight into the aetiology of diabetes. This approach provides a strong foundation for future study.