eprintid: 34415
rev_number: 16
userid: 3707
dir: disk0/00/03/44/15
datestamp: 2018-05-02 14:00:24
lastmod: 2019-05-02 05:00:02
status_changed: 2019-05-02 05:00:02
type: thesis_degree
metadata_visibility: show
contact_email: daniel.m.spagnolo@gmail.com
eprint_status: archive
creators_name: Spagnolo, Daniel M
creators_email: daniel.m.spagnolo@gmail.com
creators_id: dms167
title: Spatial statistics from hyperplexed immunofluorescence images: to elucidate tumor microenvironment, to characterize intratumor heterogeneity, and to predict metastatic potential
ispublished: unpub
divisions: sch_med_Computational_Systems_Biology
full_text_status: public
keywords: computational pathology, computational biology, multiplexed immunofluorescence, tumor microenvironment, tumor heterogeneity, pointwise mutual information,
abstract: The composition of the tumor microenvironment (TME)–the malignant, immune, and stromal cells implicated in tumor biology as well as the extracellular matrix and noncellular elements–and the spatial relationships between its constituents are important diagnostic biomarkers for cancer progression, proliferation, and therapeutic response. In this thesis, we develop methods to quantify spatial intratumor heterogeneity (ITH). We apply a novel pattern recognition framework to phenotype cells, encode spatial information, and calculate pairwise association statistics between cell phenotypes in the tumor using pointwise mutual information. These association statistics are summarized in a heterogeneity map, used to compare and contrast cancer subtypes and identify interaction motifs that may underlie signaling pathways and functional heterogeneity.
Additionally, we test the prognostic power of spatial protein expression and association profiles for predicting clinical cancer staging and recurrence, using multivariate modeling techniques. By demonstrating the relationship between spatial ITH and outcome, we advocate this method as a novel source of information for cancer diagnostics. To this end, we have released an open-source analysis and visualization platform, THRIVE (Tumor Heterogeneity Research Image Visualization Environment), to segment and quantify multiplexed imaging samples, and assess underlying heterogeneity of those samples. The quantification of spatial ITH will uncover key spatial interactions, which contribute to disease proliferation and progression, and may confer metastatic potential in the primary neoplasm.
date: 2018-05-02
date_type: published
pages: 122
institution: University of Pittsburgh
refereed: TRUE
etdcommittee_type: committee_chair
etdcommittee_type: thesis_advisor
etdcommittee_type: thesis_advisor
etdcommittee_type: committee_member
etdcommittee_type: committee_member
etdcommittee_name: Lee, Adrian V
etdcommittee_name: Chennubhotla, S Chakra
etdcommittee_name: Taylor, D Lansing
etdcommittee_name: Yang, Ge
etdcommittee_name: Fine, Jeffrey
etdcommittee_email: leeav@upmc.edu
etdcommittee_email: chakracs@pitt.edu
etdcommittee_email: dltaylor@pitt.edu
etdcommittee_email: geyang@andrew.cmu.edu
etdcommittee_email: finejl@upmc.edu
etd_defense_date: 2018-02-02
etd_approval_date: 2018-05-02
etd_submission_date: 2018-04-25
etd_release_date: 2018-05-02
etd_access_restriction: immediate
etd_patent_pending: TRUE
thesis_type: dissertation
degree: PhD
citation:   Spagnolo, Daniel M  (2018) Spatial statistics from hyperplexed immunofluorescence images: to elucidate tumor microenvironment, to characterize intratumor heterogeneity, and to predict metastatic potential.  Doctoral Dissertation, University of Pittsburgh.    (Unpublished)  
document_url: http://d-scholarship-dev.library.pitt.edu/34415/1/spagnolo-etd.pdf