Functional Annotation of Regulatory Pathways



Standardized annotations of biomolecules in interaction networks
(Gene Ontology) target function of individual molecules. Extending
such annotations to pathways is a critical component of functional
characterization of cellular signaling at the systems level.
We propose a framework for projecting gene regulatory networks
onto the space of functional attributes using multigraph models, with
the objective of deriving statistically significant pathway annotations.
We formalize the problem of identifying statistically over-represented
pathways and establish the hardness of this problem by demonstrating
the non-monotonicity of common statistical significance measures.
We propose a statistical model that emphasizes the modularity of a pathway,
and complement the statistical model by an efficient algorithm for
computing significant pathways in large regulatory networks.
Comprehensive results from our methods on the E. coli transcription
network demonstrate that the our approach is effective
in identifying known, as well as novel biological pathway annotations.

An important aspect of annotating pathways is the derivation of
quantitative metrics for distance between sets of annotations. In the
second part of the talk, we establish basic intuitive characteristics of
admissible measures of functional coherence, and demonstrate that existing,
well-accepted measures are ill-suited to comparative analyses
involving different entities (domains vs. proteins). We propose a
statistically motivated functional similarity measure that takes into
account functional specificity as well as the distribution of functional
attributes across entity groups to assess functional similarity
in a statistically meaningful and biologically interpretable manner.
We comprehensively validate our metric on detailed comparisons of
PPIs and DDIs.



--
This is joint work with Jayesh Pandey (Purdue), Mehmet Koyuturk (Case
Western Reserve), and Shankar Subramaniam (UCSD).