Organization and Function Acquisition in Protein-Protein Interaction Networks
Protein-protein interaction (PPI) networks enable the transmission of biological in-
formation throughout cells, allowing cells to respond to environmental stimuli. PPI
networks can be represented as graphs, and graph analysis techniques have been ap-
plied in order to determine the topological roles played by individual proteins in PPI
network structure. However, more complex analysis is needed to study the functional
organization of PPI networks. In addition, the proteins that make up PPI networks
change and evolve new functions over time.
In the first part of this thesis, we introduce a metric, functional insularity, to
measure the degree to which proteins physically interact with functionally related
proteins. Proteins in PPI networks exhibit significant variation in insularity values,
suggesting the presence of a tradeo between network modularity and connectiv-
ity. Low-insularity proteins|those that interact with many functionally unrelated
proteins|are more crucial than high-insularity proteins to maintaining network con-
nectivity, are less likely to be essential, and have more regulators. Furthermore, we
show that between-species homologs tend to have similar levels of functional insu-
larity. Low-insularity proteins are found between topological network modules as
well as within them. We find that functional and topological network modules con-
tain proteins with a range of insularity values, including low-insularity proteins that
might may function as \interfaces" to other modules. Finally, we show how functional
insularity analysis can be applied to improve network clustering analyses.
In the second part of this thesis, we study the acquisition of new functions by pro-
teins and their integration into the PPI network. We first use a maximum parsimony-
based approach to infer the ages of human proteins. We then determine various
function-related traits for each age group, such as protein-protein interaction count,
expression ubiquity, and number of unique domains. We find that young proteins
in human have fewer protein-protein interactions, have fewer unique domains, are
expressed in fewer tissues, and are less likely to be essential than older proteins. In
addition, we nd that proteins tend to physically interact mainly with other proteins
of similar age. Finally, we nd that younger pairs of paralogs are more coexpressed
and share more common regulators than older pairs.
In sum, this thesis advances our understanding of PPI networks by showing that
the dual requirements of modularity and connectivity are balanced using \connector"
proteins and "module" proteins, which have distinct biological traits, and by un-
covering differences between young and old proteins that suggest that proteins gain
functions and integrate into networks over time.