139
Modeling of Biomolecular Systems Interactions, Dynamics, and Allostery Poster Session II
86-POS
Board 39
Revealing Clinically Relevant Targets across Various Glioblastoma Tumor Lines by
Integrating Multiple Omic Data
Nurcan Tuncbag
1,2
, Jenny Pokorny
3
, Pamela Milani
1
, Jann Sarkaria
3
, Ernest Fraenkel
1
.
1
Massachusetts Institute of Technology, Cambridge, MA, USA,
3
Mayo Clinic College of
Medicine, Rochester, MN, USA,
4
Massachusetts Institute of Technology, Cambridge, MA,
USA.
2
Middle East Technical University, Ankara, Turkey,
With the help of high-throughput technologies, we are able to comprehensively monitor
quantitative molecular changes within signaling networks at many levels in response to
perturbations or disease. Although a single omic dataset provides a wealth of information about a
given biological problem, it is limited in the ability to fully capture a cellular pathway that is
altered in a given experimental setting. Therefore, integrating multiple data sources together in a
network context is crucial to find the omissions in each set, explore hidden entities and identify
therapeutic targets. In this work, we reconstruct signaling networks across eight implemented
xenograft models of Glioblastoma multiforme (GBM, the most aggressive type of malignant
brain tumors) patients by integrating phosphoproteomic and interactome data. The prize-
collecting Steiner forest (PCSF) algorithm has been used to reconstruct networks by taking into
account confidence that the phosphoproteomic hits are significant and the probability that each
reported protein-protein interaction is real. Simultaneous comparison of reconstructed networks
led us to find out several common proteins and pathways across multiple tumor lines that were
not identified in the phosphoproteomic dataset. For example, Nfkb signaling and mTOR
pathways were common in most of the patients, as were proteins such as Mek1, Stat6, Erbb3. We
have experimentally shown that Mek1 is stably over-phosphorylated in all tumor lines compared
to normal human astrocytes. Another target, Numb, was found to be an invasiveness marker in
this setup. Tumor lines having Numb present in their networks were in minimally invasive class
while remaining were in highly invasive class. Overall, despite the heterogeneity across various
tumor lines of GBM, common pathways and proteins revealed here have important outcomes in
GBM and our integrative network modeling approach help in finding testable therapeutic targets.
