86
F. Cheng
If the centre X or centre Y is not unique, all the nodes in centre X or centre Y are used
to define the center, and shortest path lengths between these nodes are averaged. If
the centre Y is not unique, all nodes are used to define the center and the shortest path
lengths to these nodes are averaged.
Finally, the significance of the measure is evaluated by comparing to the reference distance distribution corresponding to the expected network topological distance
between two randomly selected groups of proteins matched to size and degree (connectivity) distribution as the original disease proteins and drug targets in the human
interactome. This procedure was repeated 1000 times. As illustrated in Fig. 5.1a,
for closest distance measure (cd in Eq. 5.2), the mean distance (cd) and standard
deviation (σ cd ) of the reference distribution are used to calculate a z-score (z) by
converting an observed closest distance to a normalized distance using Eq. (5.7).
z cd =
cd − cd
σ cd
(5.7)
A detailed description for z-score calculation can be found in our recent study
[48].
5.2.6 Network Visualization and Statistical Analysis
Networks can be analyzed and visualized by Cytoscape (v3.2.0, http://www.
cytoscape.org/) and Gephi (v0.9.2, https://gephi.org/). Statistical analysis can be performed by the Python (v3.2, http://www.python.org/) or R platforms (v3.01, http://
www.r-project.org/).
5.3 Results/Case Studies
The basis for the proposed network-based methodology in this chapter rests on the
notion that the proteins that associate with and functionally govern a disease phenotype are localized in the corresponding disease module or subnetwork (graph) within
the comprehensive human protein-protein interactome network [48]. As shown in
Fig. 5.2, our preliminary network analysis reveals that cardiomyopathy-associated
proteins form a statistically, significantly clustered, distinct module in the human
protein-protein interactome, as we have previously shown for 23 other types of cardiovascular outcomes as well [48].
To examine drug effects on the cardiovascular system, our previous study [48]
quantifies the interplay between diseases and drug targets on the human proteinprotein interactome using a network proximity measure (Fig. 5.1b). Figure 5.3 reveals
the globally predicted drug-disease network using z-score (z) <−4.0, which connects
23 CV events and approximately 600 FDA-approved non-CVD drugs grouped by
F. Cheng
If the centre X or centre Y is not unique, all the nodes in centre X or centre Y are used
to define the center, and shortest path lengths between these nodes are averaged. If
the centre Y is not unique, all nodes are used to define the center and the shortest path
lengths to these nodes are averaged.
Finally, the significance of the measure is evaluated by comparing to the reference distance distribution corresponding to the expected network topological distance
between two randomly selected groups of proteins matched to size and degree (connectivity) distribution as the original disease proteins and drug targets in the human
interactome. This procedure was repeated 1000 times. As illustrated in Fig. 5.1a,
for closest distance measure (cd in Eq. 5.2), the mean distance (cd) and standard
deviation (σ cd ) of the reference distribution are used to calculate a z-score (z) by
converting an observed closest distance to a normalized distance using Eq. (5.7).
z cd =
cd − cd
σ cd
(5.7)
A detailed description for z-score calculation can be found in our recent study
[48].
5.2.6 Network Visualization and Statistical Analysis
Networks can be analyzed and visualized by Cytoscape (v3.2.0, http://www.
cytoscape.org/) and Gephi (v0.9.2, https://gephi.org/). Statistical analysis can be performed by the Python (v3.2, http://www.python.org/) or R platforms (v3.01, http://
www.r-project.org/).
5.3 Results/Case Studies
The basis for the proposed network-based methodology in this chapter rests on the
notion that the proteins that associate with and functionally govern a disease phenotype are localized in the corresponding disease module or subnetwork (graph) within
the comprehensive human protein-protein interactome network [48]. As shown in
Fig. 5.2, our preliminary network analysis reveals that cardiomyopathy-associated
proteins form a statistically, significantly clustered, distinct module in the human
protein-protein interactome, as we have previously shown for 23 other types of cardiovascular outcomes as well [48].
To examine drug effects on the cardiovascular system, our previous study [48]
quantifies the interplay between diseases and drug targets on the human proteinprotein interactome using a network proximity measure (Fig. 5.1b). Figure 5.3 reveals
the globally predicted drug-disease network using z-score (z) <−4.0, which connects
23 CV events and approximately 600 FDA-approved non-CVD drugs grouped by