Exploring relationships between multivariate radiological phenotypes and genetic features: A case-study in Glioblastoma using the Cancer Genome Atlas

Glioblastoma is a form of brain cancer with extremely poor prognosis. While comprehensive genomic profiling is routinely done to identify genetic determinants of pathological grade, a finer-course evaluation of genetic determinants of radiology-specific phenotype remains to be done. This is essential since radiological characterization is a key component of GBM diagnosis in the clinic. In this work, we seek to understand the relationship between genetic features (miRNA and mRNA) with radio-phenotypic features associated with GBM progression. Using genomics data from the Cancer Genome atlas (TCGA) as well as image-derived phenotypes from the Cancer Imaging archive (TCIA), we investigate a multi-task lasso framework to discover associations between gene expression and multivariate image phenotypes. Our study reveals that such integrated imaging-genomic analysis implicates several key molecules involved in glioma biology.