consideration all residues with a side chain of any type. This is called
the center of mass PSN (cmPSN) and relies on the identification of
contacts between side chains when the distance between their
center of mass is below a certain threshold. The calculation of the
cmPSN works in the same way as for hydrophobic contacts, except
that the calculation is extended to any residue except glycine. This is
possible as PyInteraph allows to specify which residues the hydrophobic contact interactions should be calculated among, potentially including non-hydrophobic residues as well. Especially when
running this type of analysis, the distance cut-off between centers of
mass for the definition of a contact has a major influence on the
resulting cmPSN topology. We used a 5 A ˚ cut-off, as suggested by
Salamanca Viloria et al. [25], which has been determined to be a
good value in the context of MD simulation with atomistic force
fields; this might be different depending on your setup and system
(see Note 10). In order to generate and analyze the cmPSN, we
carried out the following steps:
1. Calculate the cmPSN. We ran the pyinteraph command with
distance cut-off equal to 5 A ˚ specifying all the residue types
except glycine for hydrophobic contacts (option --hc-residues).
We specified the files containing the topology of the system and
the trajectory with the -s and -t options, respectively, and the
output file with the --hc-graph option. The latter contains the
adjacency matrix representing the cmPSN. We set the software
to use masses from the CHARMM27 force field for the reasons
described above (see Note 5) using the -ff-masses option. We
chose not to set any persistence cut-off for the interactions in
this phase, as per Note 8.
2. Filter the cmPSN to remove transient interactions. Similarly as
done before with the interaction network, we filtered the adjacency matrix file to remove the most transient interactions,
which means edges with low persistence values, using the filter_graph program. This requires as input the unfiltered
cmPSN generated with the previous command (Àd option),
and generates an output file with the same format, whose name
is specified with the -o option. We specified the persistence
threshold using the -t option. This threshold represents the
minimum percentage of frames in which the interaction must
be present over the whole trajectory. We set this threshold to
20, as previously done in previous benchmarking works [25].
3. Graph analysis. We then carried out basic graph analysis on the
resulting graphs, using the graph_analysis program, as detailed
above. We used the filtered cmPSN outputted by filter_graph as
input and considered any residue connected with 3 or more
other residues to be a hub node by setting the -k option to
3, which sets the minimum hubs degree. Similarly,
166
Matteo Lambrughi et al.
the center of mass PSN (cmPSN) and relies on the identification of
contacts between side chains when the distance between their
center of mass is below a certain threshold. The calculation of the
cmPSN works in the same way as for hydrophobic contacts, except
that the calculation is extended to any residue except glycine. This is
possible as PyInteraph allows to specify which residues the hydrophobic contact interactions should be calculated among, potentially including non-hydrophobic residues as well. Especially when
running this type of analysis, the distance cut-off between centers of
mass for the definition of a contact has a major influence on the
resulting cmPSN topology. We used a 5 A ˚ cut-off, as suggested by
Salamanca Viloria et al. [25], which has been determined to be a
good value in the context of MD simulation with atomistic force
fields; this might be different depending on your setup and system
(see Note 10). In order to generate and analyze the cmPSN, we
carried out the following steps:
1. Calculate the cmPSN. We ran the pyinteraph command with
distance cut-off equal to 5 A ˚ specifying all the residue types
except glycine for hydrophobic contacts (option --hc-residues).
We specified the files containing the topology of the system and
the trajectory with the -s and -t options, respectively, and the
output file with the --hc-graph option. The latter contains the
adjacency matrix representing the cmPSN. We set the software
to use masses from the CHARMM27 force field for the reasons
described above (see Note 5) using the -ff-masses option. We
chose not to set any persistence cut-off for the interactions in
this phase, as per Note 8.
2. Filter the cmPSN to remove transient interactions. Similarly as
done before with the interaction network, we filtered the adjacency matrix file to remove the most transient interactions,
which means edges with low persistence values, using the filter_graph program. This requires as input the unfiltered
cmPSN generated with the previous command (Àd option),
and generates an output file with the same format, whose name
is specified with the -o option. We specified the persistence
threshold using the -t option. This threshold represents the
minimum percentage of frames in which the interaction must
be present over the whole trajectory. We set this threshold to
20, as previously done in previous benchmarking works [25].
3. Graph analysis. We then carried out basic graph analysis on the
resulting graphs, using the graph_analysis program, as detailed
above. We used the filtered cmPSN outputted by filter_graph as
input and considered any residue connected with 3 or more
other residues to be a hub node by setting the -k option to
3, which sets the minimum hubs degree. Similarly,
166
Matteo Lambrughi et al.