Figure 1.
Input and output for our subnetwork inference approach.
(A) The inputs to our subnetwork inference approach are phenotypes measured in a loss-of-function assay and a background network characterizing known interactions. (B) The network elements represented in panels A, C, and other figures. (C) An inferred subnetwork for the given inputs. The subnetwork includes a directed, consistent path linking each hit (gene with an up or down phenotype) to the virus. The red borders on the unassayed nodes G and H indicate that they are inferred to have the down phenotype. Edges shown in gray are not included in the subnetwork.
Table 1.
Phenotype labels for suppressed host genes.
Table 2.
Types of host factors represented by nodes in the background network.
Table 3.
Intracellular interactions in the background network.
Table 4.
Interactions from literature.
Figure 2.
The steps of our subnetwork inference approach.
Each edge is shown with a numeric identifier for cross-reference. (A) Add a new node to the background network, representing the virus. Add connections between all nodes except no-effects to the new virus node, representing the possibility of any host factor having a direct interaction with a viral component. (B) For each hit identified by the genome-wide mutant assay, enumerate candidate paths through the background network that could explain it by providing a linear path to the virus node. (C) Infer an ensemble of consistent subnetworks. Each subnetwork is a union of paths that accounts for all of the hits and is consistent with virus phenotype data.
Figure 3.
Variables for pathway 9 from Figure 2.
The values of some variables are fixed by the data. The values of free variables are determined by the IP.
Table 5.
Integer program variables.
Figure 4.
Precision-recall curves for the hit-prediction task.
BMV at left, FHV at right. The horizontal line shows precision that would be achieved if all test cases were called hits. (A) Comparison of the diffusion kernel method to the naïve baselines. (B) Comparison of our IP approach to the diffusion kernel method and to random permutations. (C) The effect of varying , the maximum number of interfaces allowed in the subnetwork inferred by the IP method.
Figure 5.
Accuracy-coverage curves for the sign-prediction task.
BMV on the left, FHV on the right. The horizontal line indicates the accuracy that would be achieved by assigning the plurality phenotype label to every test case (down for BMV, up for FHV.)
Figure 6.
Precision-recall curves for two other objective functions on the hit-prediction task.
Comparison of this work's objective function, which maximizes node score (IP), to two alternatives inspired by published methods: maximize path count (MP-Count) and maximize path score (MP-Score). For BMV, the number of interfaces . For FHV,
.
Figure 7.
Accuracy-coverage curves for the SPINE heuristic on the sign-prediction task.
Comparison of of this work's edge sign heuristic, IP, , to the heuristic used by the SPINE method[12], IP, SPINE. Also shown is the result for our IP when
.
Figure 8.
A component from the inferred subnetwork ensemble showing the predicted involvement of Snf7p and Vps4p in viral replication.
For predictions made about node and edge relevance, confidence values <1.0 are indicated. For the unassayed nodes, the same phenotype label prediction was made in all solutions in which they appear; similarly, all solutions predicted the same direction for the undirected edges. Dashed edges indicate cases in which the edge's direction was not fixed in the background network. See Figure 1 for a key to the other network elements.
Table 6.
High-confidence predicted interfaces.
Figure 9.
A component from the inferred subnetwork ensemble showing a connection between Acb1 and the literature-extracted ubiquitin-proteasome-system interactions.
All node and edge predictions shown have confidence = 1.0 in the ensemble. A dashed edge with no terminal indicates connections to the rest of the subnetwork. Edges extracted from literature are colored blue. Doubled blue edges (as from Rsp5p to Spt23p) indicate literature-extracted edges that were also present in the original background network. See Figure 1 for a key to the other network elements.
Figure 10.
A component from the inferred subnetwork ensemble showing a connection between the literature-identified interface Ydj1p and two hits, Hsf1p and Ure2p.
The blue edge from Ydj1p to the virus was originally extracted from literature. See Figure 1 for a key to the other network elements.
Table 7.
Gene Ontology terms represented by both experimental and predicted BMV hits.
Table 8.
Additional Gene Ontology terms represented by the inferred BMV subnetwork.