structures. Edge weight is often used to quantify the relation so that
only edges that are associated with significant weights are retained;
however, the decision of connecting two residues with an edge does
not necessarily depend on the weight itself, and the two can be
independent.
The most popular edge or weight definitions that have been
employed so far are based, among others, on atomic side-chain
contacts or otherwise defined interatomic distances, on interaction
energy based on a force-field or knowledge-based potentials and
others. Other network representations of protein dynamics
are based, for instance, on local changes in the protein structure
[14–16].
1.3 Analysis of PSNs
Once a PSN has been calculated from a protein structure or ensemble, different graph analysis techniques and network parameters can
be calculated on the graph to extract useful information. One of the
simplest and most important to calculate is the degree of each node,
equal to the number of edges that are connected to that node. This
allows identifying hubs, i.e. residues that are particularly well
connected in the protein structure and are usually important for
protein stability and as communication hubs [9].
Paths are successions of residues or edges which allow reaching
a target residue from a source one. In this way, a chain of contacting
residues through which structural communication may happen can
be identified. Shortest paths are usually considered in PSNs as the
most straightforward routes of structural communication [9].
Clusters or connected components are related with the global
structure of the network and are defined as subgraphs in which
paths exist between each pair of nodes but not with the rest of the
network, representing more interconnected regions of the
protein [9].
1.4 PyInteraph
Most definitions of PSNs rely on distances between atoms, the
most popular ones being based on simple atomic contacts. While
atomic contacts are indeed descriptive of the intraprotein interactions in an ensemble, they do not account for the different physicochemical properties of the protein amino acids and do not take in
consideration specific types of non-covalent interactions that are
known to be important for stability and dynamics. Such interactions are sometimes residue-specific (for instance in the case of salt
bridges), meaning that their analysis can help in understanding how
mutations affect the interaction network in the protein. Under
these premises, a PSN based on different classes of the most relevant non-covalent interactions found in proteins may help us get a
clearer picture of the interactions occurring among residues and
complement more standard PSN analyses based on atomic contacts. This was the drive behind the development of PyInteraph, a
set of tools designed to facilitate the generation and analysis of
Interaction Networks with PyInteraph
155
only edges that are associated with significant weights are retained;
however, the decision of connecting two residues with an edge does
not necessarily depend on the weight itself, and the two can be
independent.
The most popular edge or weight definitions that have been
employed so far are based, among others, on atomic side-chain
contacts or otherwise defined interatomic distances, on interaction
energy based on a force-field or knowledge-based potentials and
others. Other network representations of protein dynamics
are based, for instance, on local changes in the protein structure
[14–16].
1.3 Analysis of PSNs
Once a PSN has been calculated from a protein structure or ensemble, different graph analysis techniques and network parameters can
be calculated on the graph to extract useful information. One of the
simplest and most important to calculate is the degree of each node,
equal to the number of edges that are connected to that node. This
allows identifying hubs, i.e. residues that are particularly well
connected in the protein structure and are usually important for
protein stability and as communication hubs [9].
Paths are successions of residues or edges which allow reaching
a target residue from a source one. In this way, a chain of contacting
residues through which structural communication may happen can
be identified. Shortest paths are usually considered in PSNs as the
most straightforward routes of structural communication [9].
Clusters or connected components are related with the global
structure of the network and are defined as subgraphs in which
paths exist between each pair of nodes but not with the rest of the
network, representing more interconnected regions of the
protein [9].
1.4 PyInteraph
Most definitions of PSNs rely on distances between atoms, the
most popular ones being based on simple atomic contacts. While
atomic contacts are indeed descriptive of the intraprotein interactions in an ensemble, they do not account for the different physicochemical properties of the protein amino acids and do not take in
consideration specific types of non-covalent interactions that are
known to be important for stability and dynamics. Such interactions are sometimes residue-specific (for instance in the case of salt
bridges), meaning that their analysis can help in understanding how
mutations affect the interaction network in the protein. Under
these premises, a PSN based on different classes of the most relevant non-covalent interactions found in proteins may help us get a
clearer picture of the interactions occurring among residues and
complement more standard PSN analyses based on atomic contacts. This was the drive behind the development of PyInteraph, a
set of tools designed to facilitate the generation and analysis of
Interaction Networks with PyInteraph
155