Abstract

The practice of personalized medicine requires consideration of multiple individual modalities as well as their principled integration. Considering a problem in predictive oncology, we outline our groups work integrating histology, radiology and genomics data from the TCGA brain datasets. The first problem, pertaining to radiogenomics, aims to derive radiological surrogates of key genetic alterations, with an eye towards non-invasive assessment. The second problem visits histopathology data to derive imaging surrogates that are genomic ally relevant. The third problem is a investigation into spatial profiling datasets coming from multiplex IF and spatial transcriptomics and studies problems in spatially informed network inference.