Milenkovic, tmilenko AT nd DOT edu
Introduction: Initial protein structural comparisons were sequence-based. Since amino acids that are distant in the sequence can be
close in the 3-dimensional (3D) structure, 3D contact approaches can complement sequence approaches. Traditional 3D
contact approaches study 3D structures directly and are alignment-based. Instead, 3D structures can be modeled as protein
structure networks (PSNs). Then, network approaches can compare proteins by comparing their PSNs. These can be
alignment-based or alignment-free. We focus on the latter. Existing network alignment-free approaches have drawbacks:
1) They rely on naive measures of network topology. 2) They are not robust to PSN size. They cannot integrate 3) multiple
PSN measures or 4) PSN data with sequence data, although this could improve comparison because the different data types
capture complementary aspects of the protein structure. We address this by: 1) exploiting well-established graphlet measures
via a new network alignment-free approach, 2) introducing normalized graphlet measures to remove the bias of PSN size, 3)
allowing for integrating multiple PSN measures, and 4) using ordered graphlets to combine the complementary PSN data and
sequence (specifically, residue order) data. We compare synthetic networks and real-world PSNs more accurately and faster
than existing network (alignment-free and alignment-based), 3D contact, or sequence approaches.
Reference: Fazle E. Faisal, Khalique Newaz, Julie L. Chaney, Jun Li, Scott J. Emrich, Patricia
L. Clark, and Tijana Milenkovic (2017), GRAFENE: Graphlet-based alignment-free network
approach integrates 3D structural and sequence
(residue order) data to improve protein structural
Software: Our Unix version implementation for
performing protein structural comparison is available here.
Usage: ./psn-classify [contact-map-pdb-dir]
[domain-annotation-file] [psn-approach] [output-dir] -c [cut-off] -v
[node-threshold] -d [max-distance-threshold] -o [component-threshold] -s
[group-elements-threshold] -r [variation-threshold] -k [long-range]
[contact-map-pdb-dir] contains a set of files, where
each file stores contact map information (with respect to three
distance cut-offs: 4A, 5A, and 6A) for a protein. Protein contact maps
used in our study are available here.
[protein-annotation-file] is the file, which
contains the annotation information of each protein in [contact-map-pdb-dir].
Protein annotation files used in our study are available here.
[psn-approach] is the protein structural
comparison approach. Our software supports 17 approaches: a)
Graphlet-3-4, b) Graphlet-3-5, c) OrderedGraphlet-3, d) OrderedGraphlet-3-4,
NormGraphlet-3-4, f) NormGraphlet-3-5, g) NormOrderedGraphlet-3, h) NormOrderedGraphlet-3-4,
Average-degree, j) Average-distance, k) Maximum-distance, l)
Average-closeness-centrality, m) Average-clustering-coefficient, n)
Intra-hub-connectivity, o) Assortativity, p) Existing-all, and q) Sequence.
[output-dir] is the directory that contains the
output of the comparison.
[cut-off] is the distance cut-off. The software uses
a distance cut-off 4A by default. Other options for the cut-off are 5A and 6A.
[node-threshold] is the least number of nodes in a
PSN to be considered for comparison. The software uses a threshold
0 by default. Our suggested value for this parameter is 101.
[max-distance-threshold] is the least maximum
distance in a PSN to be considered for comparison. The software
uses a threshold 0 by default. Our suggested value for this parameter
[component-threshold] is the maximum number of
components in a PSN to be considered for comparison. The software
allows unlimited number of components in a PSN by default. Our
suggested value for this parameter is 1.
[group-elements-threshold] is the minimum number of
PSNs that belong to a class (or group) according to the
[protein-annotation-file]. Our suggested value for this parameter is
[variation-threshold] is the least possible
variation in the input data as captured by the first r
principal components resulted from principal component analysis. Our
suggested value for this parameter is 0.90.
[long-range] is the "long-range(K)" constraint to be
applied while counting ordered graphlets. The software uses a default
value of 1 for this parameter.
Example: Given contact maps of proteins in the
directory "ContactAllPDB" and domain annotations in the file "cath-primary.txt",
protein structural comparison can be computed by the PSN-approach
"NormOrderedGraphlet-3-4" with the following command.
./psn-classify ContactAllPDB cath-primary.txt
NormOrderedGraphlet-3-4 output-dir -c 4A -v 101 -d 6 -o 1 -s 30 -r 0.90 -k 2
The command will perform protein structural comparison
using NormOrderedGraphlet-3-4 on the PSNs that are annotated by CATH primary
classes. The command ensures that PSNs having at least 101 nodes, at least
six maximum distance, and at most one component are considered for
comparison. The command also ensures that a CATH primary class is
considered only if there are at least 30 PSNs within the class (or group).
Additionally, the command ensures that we only count the graphlets in which
every pair of interacting nodes (amino acids) are at least 2 distance apart in the
protein sequence. The comparison will be done by performing principal component analysis
and by selecting first r of the resultant principal components that
account for at least 90% variation in the input data.
The output file
output-dir/pr-roc-normorderedgraphlet-3-4.txt contains comparison
outcomes according to precision, recall, area under precision recall (AUPR),
and are under receiver operator characteristic curve (AUROC).