INTEGRATING MULTI-OMICS AND EHR DATA FOR PERSONALIZED DISEASE PREDICTION USING DEEP LEARNING IN DATA WAREHOUSING AND MINING ENVIRONMENTS: A REVIEW

Abstract

The advent of precision medicine has necessitated the integration of heterogeneous biomedical data sources to unravel the complex mechanisms underlying human diseases. While high-throughput technologies have generated vast amounts of multi-omics data (genomics, transcriptomics, proteomics) and Electronic Health Records (EHRs) provide rich phenotypic information, the effective fusion of these modalities remains a significant challenge. This review paper critically analyses the current state of "Integrating Multi-Omics and EHR Data" for personalized disease prediction, with a specific focus on Deep Learning (DL) methodologies within Data Warehousing and Mining frameworks. We examine recent advancements in data fusion strategies Early, Intermediate, and Late and evaluate the efficacy of deep neural architectures, including Multi-modal Autoencoders, Graph Convolutional Networks (GCNs), and Transformer-based models. Furthermore, the review identifies critical gaps in current data warehousing infrastructures regarding their ability to handle the high dimensionality and sparsity of omics data alongside the unstructured nature of clinical notes. By synthesizing findings from recent high-impact literature (2020–2025), we propose a unified, scalable framework that leverages advanced data mining techniques to bridge the gap between molecular biology and clinical informatics for accurate, real-time personalized healthcare.