Abstract:

Acquiring a systematic understanding of the operation of biological  
networks constitutes a major research goal in the field of  
Computational Biology. In this talk, I will describe two strands of my  
research, concerning the function of metabolic and protein-protein  
interaction (PPI) networks, without assuming any deep prior biological  
knowledge. The talk will provide an overview of the basic  
computational ideas, and the main biological insights emerging.

(i) Current research of PPI networks focuses on analyzing the static  
topology of the network and dissecting it into functional modules, an  
essential step towards the understanding of its operation. For this  
purpose, we developed network query algorithms which find modules that  
are conserved across multiple species, by efficiently searching for  
homeomorphic subgraphs using the probabilistic color-coding technique.  
These methods enabled us to predict a large number of previously  
uncharacterized signaling pathways and protein complexes conserved  
across yeast, fly and human.

(ii) In metabolic networks, basic physicochemical principles enable  
the large-scale modeling of complex behaviors via a constraint-based  
modeling approach. Within this framework we developed computational  
methods, based on Mixed Integer Linear Programming (MILP), for  
predicting metabolic phenotypes under various environmental and  
genetic conditions. These methods were used to explore the complex  
interplay between transcriptional regulation and metabolism, and to  
produce a new systematic functional annotation of metabolic genes.  
Employing a recently published network of human metabolism, we  
successfully predicted tissue-specific metabolic behavior of  
disease-causing genes, opening the way towards the computational study  
of many metabolic disorders.