Integrative Analysis of Multimodal Omics Data

With advancements in technology and the decreasing cost of data acquisition, high-throughput omics data have become increasingly prevalent in biomedical research. These data are often collected across multiple omics modalities at different molecular levels, offering a comprehensive perspective on underlying biological mechanisms. However, the multimodal nature of multiomics data presents unique and complex challenges for statistical analysis. In this article, we provide a comprehensive review of recent advancements in statistical methods for multiomics data integration. We discuss key topics in unsupervised learning (including dimension reduction, clustering, and network analysis), supervised learning (including regression, classification, and mediation analysis), and other areas. Finally, we highlight unresolved challenges and propose promising directions for future research to further advance the field.