Bud Mishra

NYU

Title: Pathways and Abstraction

Ilya Korsunsky, Andreas Witzel, Andrew Sundstrom, Bud Mishra


Abstract: Cell autonomous models in cancer usually consist of integrated
pathway models of intracellular signaling. These integrated networks are
useful for understanding timing of stochastic biochemical events but
provide little ready, intuitive insight into the high-­‐level behavior
of the cell. We introduce a pathway abstraction model (PAM) that retains
the timing information of key events in a biochemical network. It
abstracts the intractable number of combinatorial networks states into a
finite number of biologically meaningful cellular states.

We represent PAM as a graph in which nodes denote phenomenological
states of the cell, and edges denote timed responses to combinations of
input. We construct PAM by sampling the network using a modified
adaptive boosting algorithm and control the exponential search space of
possible models using beam search. We successfully demonstrated this
approach on simple models, such as the bistable switch of Rb-­‐ E2F in
the control of the G1‐S checkpoint (Yao et al Nature 2008). Moreover,
we assembled and encoded models that represent the majority of the 12
core pathways of pancreatic cancer (Jones et al Science 2008) and plan
to characterize these models using PAM.

To readily understand the meaning of these states, we plan to label them
with a standard ontology (e.g. KEGG or GO). This provides a powerful
framework for thinking about cellular behavior in phenomenological
terms. As a prime example of this capability, we plan to describe
hallmarks of cancer in terms (Hanahan et al Cell 2011) of temporal logic
formulae on PAM. Using these formulae, we will model check our cell to
characterize the cell in terms of its distance to the 6 classic and 4
emerging hallmarks of cancer.