unexplored. One such intricacy is node clustering. The overall
connectivity in the network leads to node clusters, such that connectivity between nodes of same cluster is higher compared to that
across clusters. While the difference in edge weights between same
nodes due to minor perturbation in a network speaks about the
local variation in connections between them, they may sometimes
lead to a change in global clustering among the nodes in the
network. The change in node clustering can be identified by comparing Fiedler vectors of both the systems where Fiedler vector is
defined as the eigen vector corresponding to the second smallest
eigen value. Recent developments in network theory include Network Similarity Score (NSS) which considers all eigen vectors and
eigen values of two networks and compares them at various levels
like local edge weight, local clustering change, and global clustering
changes [13, 14]. The unique advantage of the method lies in
accurate scoring in the case of comparison between large number
of extremely similar networks and also in identifying the regions of
differences between the networks.
The insight regarding the changes that occur due to allostery
and protein-protein interactions is brought by studying the clustering of nodes and changes in it, which forms the primary focus of
this chapter.
3 Construction and Analysis of PSNs
Network representation of protein structures has proved to be
successful in addressing various problems related to protein folding,
dynamical behavior, ligand binding, and interactions between proteins. Network parameters such as hubs, clusters, cliques, communities, and shortest paths are the most common ones that are
evaluated for characterization of critical residues, folded and
unfolded states, and long-range communication during proteinprotein interactions. A detailed overview of analysis of various
metrics from network methodology has been discussed in reviews
[10, 11] and references therein. A succinct representation is given
here, which is a prerequisite to follow the recent developments
presented in the following section.
The three-dimensional structure of a polypeptide chain is dictated by an optimal non-covalent interaction between different
amino acids in the chain and various definitions are used to represent the non-covalent interactions in a PSN. For example, backbone networks are considered to study gross characteristics such as
domain identification and protein folding [20, 21]. Similarly, sidechain networks are considered to study clusters of residues such as
those involved in the interactions of the protein with other proteins
92
Vasundhara Gadiyaram et al.
connectivity in the network leads to node clusters, such that connectivity between nodes of same cluster is higher compared to that
across clusters. While the difference in edge weights between same
nodes due to minor perturbation in a network speaks about the
local variation in connections between them, they may sometimes
lead to a change in global clustering among the nodes in the
network. The change in node clustering can be identified by comparing Fiedler vectors of both the systems where Fiedler vector is
defined as the eigen vector corresponding to the second smallest
eigen value. Recent developments in network theory include Network Similarity Score (NSS) which considers all eigen vectors and
eigen values of two networks and compares them at various levels
like local edge weight, local clustering change, and global clustering
changes [13, 14]. The unique advantage of the method lies in
accurate scoring in the case of comparison between large number
of extremely similar networks and also in identifying the regions of
differences between the networks.
The insight regarding the changes that occur due to allostery
and protein-protein interactions is brought by studying the clustering of nodes and changes in it, which forms the primary focus of
this chapter.
3 Construction and Analysis of PSNs
Network representation of protein structures has proved to be
successful in addressing various problems related to protein folding,
dynamical behavior, ligand binding, and interactions between proteins. Network parameters such as hubs, clusters, cliques, communities, and shortest paths are the most common ones that are
evaluated for characterization of critical residues, folded and
unfolded states, and long-range communication during proteinprotein interactions. A detailed overview of analysis of various
metrics from network methodology has been discussed in reviews
[10, 11] and references therein. A succinct representation is given
here, which is a prerequisite to follow the recent developments
presented in the following section.
The three-dimensional structure of a polypeptide chain is dictated by an optimal non-covalent interaction between different
amino acids in the chain and various definitions are used to represent the non-covalent interactions in a PSN. For example, backbone networks are considered to study gross characteristics such as
domain identification and protein folding [20, 21]. Similarly, sidechain networks are considered to study clusters of residues such as
those involved in the interactions of the protein with other proteins
92
Vasundhara Gadiyaram et al.