Figure 1.
This diagram provides an outline of our method, steps are numbered according to the order in which they are discussed in the main text.
Figure 2.
An example portion of the interactions matrix used in biclustering.
(A) Shows an example portion of the interactions matrix. ‘1’ represents the presence of a given interaction, while ‘0’ the absence of that interaction, between a human protein interactant (shown left) and an HIV protein; the interaction having a given outcome (shown above). The entire matrix was biclustered to identify sets of host proteins that undergo the same set of HIV-1 interactions. (B) Shows an example bicluster that would be found in the portion of matrix given in (A).
Figure 3.
Venn diagrams showing biological cohesiveness among proteins within significant biclusters, using three measures.
Counts refer to the number of biclusters that include human proteins that are significantly biologically related () from a possible 279. (A) Displays three network clustering measures: shared edge count, average shortest path and largest connected component. (B) Displays semantic similarity in terms of the three GO ontologies. (C) Displays the overlap of all three measures of biological cohesiveness: semantic similarity, network clustering and sequence similarity.
Figure 4.
Comparison of protein pairs within significant biclusters to other protein pairs.
Panels A, B and C show the semantic distance distributions for the three GO ontologies: biological process, cellular component and molecular function, respectively, for (i) human protein pairs from significant biclusters (shown in grey) and (ii) all other human protein pairs from PBPs (shown in black). Panel D shows the pairwise sequence similarity distributions for (i) and (ii). These charts show that human proteins from within significant biclusters are more similar in their GO annotation and sequence than other protein pairs ( in a Mann-Whitney U test, in all cases).
Figure 5.
Tree showing the relationship between significant biclusters and higher-level host subsystem groupings.
Individual biclusters are represented by terminal branches. Relationships are derived using a distance measure based on the proportion of shared interactions between significant biclusters and the tree was drawn using the neighbor joining method. The tree is divided into sections that show the higher-level host subsystems, largely derived using the tree structure. Subsystems of biclusters are colour coded (see key). Biclusters not labelled are those that have been placed in a biologically related group not adjacent on the tree.
Figure 6.
These networks represent the pattern of cytokine regulation in the cytokine-activity host subsystem that were defined through identifying significant patterns of HIV-host interaction. Edges represent PPIs. Edge width is proportional to the number of PPIs being represented. For clarity, we only show PPIs that are reported more than once in the HHPID. These networks show that cytokine dysregulation due to HIV-1 infection is wide reaching and complex, affecting many host cytokines, both via upregulation (left) and downregulation (right).
Figure 7.
HIV-1-host interaction patterns, by HIV-1 protein.
This network illustrates core patterns of HIV-host interaction. The human host is depicted as a series of cellular subsystems, represented by orange circular nodes, where the diameter of the node is proportional to the number of host proteins within that subsystem. HIV-1 is depicted by the viral proteome (blue triangles). Interactions between HIV-1 proteins and host subsystems are represented by edges, where the edge width is proportional to the number of interactions. For clarity, only those interactions that are shared by over half of the host proteins in a subsystem are shown. *Indicates a host subsystem whose subsystem annotation corresponds to a statistically significant group among HDFs ().
Indicates a statistically significant intersection between the subsystem proteins and HDF set (
).
Figure 8.
HIV-1-host interaction patterns, by interaction type.
This network illustrates core patterns of HIV-host interaction. The human host is depicted as a series of cellular subsystems, represented by orange circular nodes, where the diameter of the node is proportional to the number of host proteins within that subsystem. The action that HIV-1 has on these subsystems is depicted by a series of interaction outcomes (blue diamonds). Interactions between HIV-1 and host subsystems are represented by edges where the edge width is proportional to the number of interactions. The directionality of the interaction is implicit in the description of the interaction outcome. For example, the edge linking the MHC protein complex node and the ‘upregulates’ node represents ‘HIV-1 upregulates the MHC protein complex’, whereas the edge linking the cytokine activity node and the ‘activated by’ node represents ‘HIV-1 is activated by cytokine activity’. For clarity, only those interactions that are shared by over half of the host proteins in a subsystem are shown. *Indicates a host subsystem whose subsystem annotation corresponds to a statistically significant group among HDFs ().
Indicates a statistically significant intersection between the subsystem proteins and HDF set (
).