Predictive morphometric subtypes in Glioblastoma Multiforme

The talk will begin with an overview of the research on (i) high-content screening of multicellular systems, (ii) computational histopathology, and (iii) molecular transporters. It will then focus on computational histopathology and future directions. Traditional approach to computational histopathology has focused on computer-aided pathology (CAP). In recent years, the field of computational pathology has extended to precision medicine via proliferation of large-scale data such as The Cancer Genome Atlas (TCGA). TCGA is a rapidly expanding resource that is accelerating discovery in cancer by providing the research community with mineable genomics and clinical outcome data. Pathology images, from H&E stained samples, have been added to complement the molecular and clinical data. However, utilization of whole slide images is substantially hindered by the batch effects, biological heterogeneity, tumor composition, and complexities of tumor architecture. Computational techniques will be presented that overcome these complexities to reveal intrinsic predictive subtypes from the morphometric signature of a cohort of 250 GBM patients. Molecular correlates of each subtype are then constructed for potential targeted therapy. In addition to computed morphometric subtypes, tumor heterogeneity is also examined to evaluate whether heterogeneity is more virulent in predicting the outcome.