Molecular Basis of Tumor Heterogeneity and Morphometric subtypes in Glioblastoma Multiforme (GBM)
Lawrence Berkeley National Laboratory
University of California
April 13, 2012
Fischbach Room, Folsom Library - 1:30 p.m. to 2:30 p.m.
Refreshments at 1:15 p.m.
We hypothesize that tumor histopathology reflects the interaction of underlying molecular defects and environmental factors, and that quantification of morphological data will provide a new systems biology approach for classifying tumor subtypes. We predict that computed morphometric indices will be instrumental in determining which histological aberrations affect tumor aggressiveness and treatment responsiveness, and are therefore suitable therapeutic targets. Furthermore, quantitative characterization of tumor sections reveals the inherent heterogeneity (e.g., variations in organization, cellular features) that is lost during molecular profiling. The concept of heterogeneity may be an important indicator of the tumor growth and clinical outcome.
We present a pipeline that has been implemented to process large scale tumor sections with a significant amount of technical (e.g., fixation, staining) and biological (e.g., cell type, cell signature) variations from large cohorts of The Cancer Genome Atlas (TCGA) data. The pipeline utilizes advanced techniques in image-based modeling to remove batch effects and to model tumor signatures based on their intrinsic properties. The computational pipeline has been applied to multiple TCGA tumor types for computing a multidimensional representation on a cell-by-cell basis. Such a representation can then be queried through multiple policies for identifying morphometric subtypes based on intrinsic properties of tumor signatures from a large scale data bank. We show that in certain cases, subtypes can be prognostic and/or predictive of the clinical outcomes, and that the molecular basis of each subtype provides new insights into the underlying biological processes. In a companion study, we also examine molecular basis of tumor heterogeneity in the context of disease progression.
Bahram Parvin received his Ph.D. in Electrical Engineeing from the University of Southern California in 1991. His laboratory has focused on developing technologies for (i) high content screening of multicellular systems, (ii) integrating diverse genome-wide molecular data with phenotypic data, and (iv) identifying novel probes and delivery vehicles with specific properties. He is a staff scientist with Lawrence Berkeley National Laboratory and an adjunct Professor of Electrical Engineering at the University of California.
Hosted by: Dr. Bulent Yener (x6907)
Last updated:April 10, 2012