Figure 1.
Experimental reliability of annotated PPIs.
(A) All 101,131 PPIs in HIPPIE are scored according to their associated experimental evidence with a value that ranges from 0 to 1 and increases with the quality and amount of experimental evidence reported in PPI databases [14]. We were able to infer context to a fraction of interactions according to: GO terms biological process (BP) and cellular component (CC), MeSH terms (subcategories disease and tissue) and tissue or housekeeping expression. The numbers in the bars indicate the mean experimental score of the non-annotated fraction (above, black font) and of the annotated fraction (below, white font), respectively. All mean-score differences between annotated and not annotated interactions were significant (p<0.001; Mann-Whitney-test). (B–C) Box plots visualizing the distribution of experimental scores of PPIs associated with GO (left) and MeSH (right) term categories. (B) The scores for GO and MeSH terms decreased generally for less specific terms (the only exception was GO terms depth 2, which was associated with interactions of a lower mean confidence as compared to GO terms depth 1). (C) GO and MeSH terms were subdivided in quartiles according to the number of interactions annotated for each category. The scores decreased for terms associated to higher numbers of interactions.
Figure 2.
Tissue-specific PPI subnetwork of human proteins interacting with influenza virus proteins.
(A) Influenza proteins (red) interact with 23 ‘first layer’ host proteins (blue). These first layer proteins have interaction partners that are specific for the bronchial epithelial tissue (BET) subnetwork, for the lung subnetwork or are shared between both subnetworks (all in green). Details of the genes in this figure are given in a Cytoscape file (File S1). (B and C) Mini-networks for BET (B) and lung (C) were created from tissue-specific protein networks linking viral proteins to host proteins whose transcript was up-regulated after influenza virus infection (for a complete list of interactions, see Supplementary Table S4). Viral protein nodes are shown in red, first layer host interactors in blue and second layer host interactors in green. The STAT1 protein (shown in (B) as the white node) was not one of the original network-derived nodes, but was included due to its association with two other network nodes (BHLHE40 and HDAC1) and its known role in mediating inflammation in response to viral infection. General functions associated with different areas in each mini-network (e.g., ‘Inflammation’ and ‘Focal adhesions’) are described by partially transparent colored boxes in both (B) and (C).
Figure 3.
Protocol for extraction of a PPI subnetwork related to phosphorylation in Alzheimer's disease.
The flowchart illustrates the input terms and options used to generate a topic-focused PPI subnetwork. Eight genes were selected as a result of an unbiased literature mining query for proteins related to Alzheimer's disease (AD) and phosphorylation (see main text for details). The PPI network of first neighbours of these genes in HIPPIE was generated. Then, filters were applied to focus on a PPI subnetwork or proteins expressed in the brain and related to cell death, thus relevant to AD.
Figure 4.
Filtering and highlighting a PPI subnetwork related to phosphorylation in Alzheimer's disease.
A PPI network was generated as explained in Figure 3 starting with 8 genes relevant for Alzheimer's disease (AD) and phosphorylation. (A) The PPI network contains 727 interactions. (B) Filtering for interactions between partners that are housekeeping or expressed in the brain (“whole brain” and “prefrontal cortex”), relate to the GO term “cell death”, and with experimental scores above 0.5, results in a much more focused subnetwork involving 6 of the 8 genes used as input (octagonal nodes). Nodes corresponding to receptors and transcription factors are colored (blue and pink nodes, respectively). Edge directed path analysis from receptors to transcription factors resulted in the association of directionality to some of the edges (arrows). The path LRP6-GSK3B-MAPT-AATF is highlighted in green and described in the text.