An Integrative Statistical Framework for Multi-Modal Omics Data

It is becoming increasingly common for patients to be profiled across multiple molecular compartments -genomic, transcriptomic, proteomic, metab​olomic, etc. We develop a framework that leverages recent developments in the estimation of high-dimensional multi-layered graphical models that provide insights on regulatory mechanisms across molecular compartments (layers), as well as on molecular interactions within each layer and are also capable of accommodating outcome variables such as disease risk, or patient survival times. We discuss algorithmic issues, establish theoretical properties of the estimates and apply them to real data from a hyperurecimia study.