Removing bias against membrane proteins in interaction networks
Department of Biochemistry and Biomedical Sciences, McMaster University, Hamilton, Ontario L8N 3Z5, Canada
BMC Systems Biology 2011, 5:169 doi:10.1186/1752-0509-5-169Published: 19 October 2011
Additional file 1:
PIN network. PIN interactions with p-values and scores.
Format: TXT Size: 4.6MB Download file
Additional file 2:
Supplementary figures. This file contains the following Supplementary figures and table: Figure S1 - Network A interactions are distributed evenly across the top 60,000 of the 94,879 interactions in PIN. Coverage of Network A (rich in membrane interactions) is shown for interactions with different cutoffs of PIN. Figure S2 - Construction of iREF network as a control weighted network. The iREF network was constructed by the optimized combination of two networks: A' comprised of interactions not previously used from the iREF database and B, the same network B used to construct PIN. The area under the curve (AUC) of Precision-Recall plots for the top 5,000 interactions was used for selecting the optimum weight for network B using membrane proteins from GO Slim as an established benchmark dataset. The AUC was calculated for different weights for network B, and the weight corresponding to prediction performance in the plateau with higher number of membrane interactions (AUC = 0.38) was chosen as the optimal weight of network B (0.8). Like PIN the performance of this network increased with increasing membrane interactions. Figure S3 - The enrichment analysis of PIN shows over-represented gene categories (red) occur only in the lower significance quintiles. The analysis was repeated 10 times by randomly resampling 1000 interactions in each different quintile of the network (I-V). The median p-value of the 10 analysis is shown. A) Cell compartment. B) Biological process. C) Molecular function. Figure S4 - PIN increases the proportion of membrane associated interactions in the network compared to a hypergeometric model or random scoring. Coverage of interactions by PIN of genes with related GO terms. A) membrane complexes (GO:0030119: AP-type membrane coat adaptor complex [17 proteins], GO:0072546: ER membrane protein complex [6 proteins], GO:0005744: mitochondrial inner membrane presequence translocase complex [11 proteins], GO:0005742: mitochondrial outer membrane translocase complex [8 proteins], GO:0030119: AP-type membrane coat adaptor complex [17 proteins], GO:0072379: ER membrane insertion complex [7 proteins], GO:0005744: mitochondrial inner membrane presequence translocase complex [11 proteins]) or B) Transport genes Figure S5 - Correlation between Indess(k) and membrane content varies with k Indess(k) for k = 5-100 varies between -0.5 and -0.2 and is always negative. Figure S6 - Hierarchical clustering using PIN separates starvation induced genes from other autophagy process genes. The similarity between interactions was used to build a dendrogram where each leaf is a link (interaction) from the original network and branches represent link (interaction) communities. Each row or column represents an interaction in the autophagy network and the values in the matrix are Z-scored similarity distances between interactions. The red squares in the matrix represent modules as sets of closely interrelated interactions. Each square in the matrix was examined for overlap with each ATG process and annotated manually (bars to the right). Table S1 - Network density and clustering coefficient are not correlated with the fraction of membrane interactions reported by each technique.
Format: PPT Size: 558KB Download file
This file can be viewed with: Microsoft PowerPoint Viewer
Additional file 3:
GO enrichment of hubs. GO enrichment of hubs (degree > 10) for different networks.
Format: TXT Size: 11KB Download file
Additional file 4:
Cytoscape file. Cytoscape format file of autophagy and UPR interactions from PIN.
Format: CYS Size: 168KB Download file