We hypothesize that large cohorts of brain histology sections can stratify patients for improved diagnostic and therapy. However, processing a large cohort of histology sections requires advanced algorithm and software infrastructure for visualization and data analytic. We will develop and validate a computational platform for stratifying brain tumors for the applications of precision medicine. The platform will profile a large cohort of histology sections, of the brain by computing morphometric subtypes. Subsequently, morphometric subtypes will be used to enrich genomic information to reveal morphometrically enhanced genomic subsets (MEGS), and to identify epigenetically regulated genes. Morphometric indices will be computed by profiling whole slide images (WSI) of H&E stained tumor sections, which will enable multi-parametric representation in terms of nuclear- shape, -types, -organization, and aberrant regions of histopathology. However, robust processing of the H&E stained WSIs is not without challenges and suffers from batch effects, biological heterogeneity, and complexities associated with aberrant histopathology. The proposal has two aims and deliverables.
In Aim 1, we will (i) develop and refine computational methods for eliminating the batch effects and quantifying aberrations in tumor signature at multiple levels; (ii) investigate whether genomic aberrations can be classified using H&E stained sections; and (iii) represent metrics for characterizing tumor heterogeneity.
In Aim 2, morphometric subtypes and tumor heterogeneity indices that are predictive of the outcome will be identified. These morphometric subtypes and tumor heterogeneity indices will then be used to enrich genomic and epigenetic signatures for validation on independent samples using immunohistochemistry. Computational components of Aims 1-2 will be integrated with the open source software platform that is being developed for managing and visualizing histology sections by Kitware, Inc. The final product will be to stratify a new patient against an atlas of precomputed morphometric subtypes that are predictive of the outcome, label the pathology with a published genomic subtype, and to generate new hypothesis for improved targetted therapy.
A novel initiative is proposed to profile H&E stained whole slide images (WSI) for clinical application of precision medicine for adult brain tumor patients. Large cohorts of low-grade gliomas (LGG) and Glioblastoma Multiforme (GBM) will be (i) processed to generate morphometric subtypes and tumor heterogeneity indices that are predictive of survival or response to therapy; and (ii) used to investigate whether published genomic subtypes can be classified using low-cost H&E stained histology sections.