When the human viral infectome and diseasome networks collide: towards a systems biology platform for the aetiology of human diseases

Abstract

Background: Comprehensive understanding of molecular mechanisms underlying viral infection is a major challenge towards the discovery of new antiviral drugs and susceptibility factors of human diseases. New advances in the field are expected from systems-level modelling and integration of the incessant torrent of high-throughput "-omics" data.

Results: Here, we describe the Human Infectome protein interaction Network, a novel systems virology model of a virtual virus-infected human cell concerning 110 viruses. This in silico model was applied to comprehensively explore the molecular relationships between viruses and their associated diseases. This was done by merging virus-host and host-host physical protein-protein interactomes with the set of genes essential for viral replication and involved in human genetic diseases. This systems-level approach provides strong evidence that viral proteomes target a wide range of functional and inter-connected modules of proteins as well as highly central and bridging proteins within the human interactome. The high centrality of targeted proteins was correlated to their essentiality for viruses' lifecycle, using functional genomic RNAi data. A stealth-attack of viruses on proteins bridging cellular functions was demonstrated by simulation of cellular network perturbations, a property that could be essential in the molecular aetiology of some human diseases. Networking the Human Infectome and Diseasome unravels the connectivity of viruses to a wide range of diseases and profiled molecular basis of Hepatitis C Virus-induced diseases as well as 38 new candidate genetic predisposition factors involved in type 1 diabetes mellitus.

Conclusions: The Human Infectome and Diseasome Networks described here provide a unique gateway towards the comprehensive modelling and analysis of the systems level properties associated to viral infection as well as candidate genes potentially involved in the molecular aetiology of human diseases.

Figure 1

The Human Infectome Diseasome Newtok. a. The Human Infectome and Diseasome Network. Viral and host cellular proteins are respectively represented as red and blue nodes. Interactions between host cell proteins are represented by blue edges and define the Host Protein Interaction Network (HPIN). Interactions between viral and host proteins (red edges) are plugged onto the HPIN to define the Human Infectome Network (HIN). Host cell proteins directly interacting with viral proteins are named targeted proteins (TPs). Host cell proteins at a 1-hop distance of a targeted protein are named targeted proteins neighbours - (TP-Ns). Edges between human diseases (black nodes) and host cell proteins represent disease-gene associations as referred by OMIM. Proteins associated to OMIM diseases are termed disease-related proteins (DRPs). DRPs are either directly targeted by viruses (red stroke), or indirectly targeted by viruses (orange stroke). Network integration of HIN with the whole set of disease-gene associations defines the Human Infectome-Diseasome Network (HIDN). b. The virus-host interactome. The virus-host interactome part of HIN is represented as a multi-coloured graph (available in interactive format in Additional file 1). Blue nodes represent host cell proteins and viral proteins are colourized according to their taxonomy origin. Only the most connected viruses among the 110 viruses are represented in the legend: HPV - Human Papillomavirus; SV - Simian virus; HIV - Human Immunodeficiency Virus; HHV - Human Herpes Virus; HCV - Hepatitis C Virus; PTLV - Primate T-lymphotrophic Virus; HBV - Hepatitis B Virus. The size of nodes is proportional to the connectivity of proteins within the virus-host interactome. c. Distribution of viral protein and cellular protein connectivities in the Human Infectome Network. Probability density distributions P(k) of viral protein connectivity (k_h(VP)) and host protein connectivity values (k_h(TP)) are respectively given in red and blue and are linearized in log scale.

Figure 2

Topological properties of the Human Infectome Network . a. Topological properties of virus-targeted proteins. The average connectivity (top), centrality (middle) and bridging centrality (bottom) properties of targeted proteins (red bars) are compared to that of HPIN proteins not targeted by any virus (blue bars). Average measures are split into low connectivity proteins (LD - with a connectivity inferior or equal to 5, the median threshold) and high-connectivity proteins (HD - with connectivity superior to 5). Differences were statistically assessed using one-tailed Wilcoxon test; P-Values associated to the test are given (NS - Non significant testing P-Value > 0.05; P-Value < 0.05 *; P-Value < 0.01 **). b. The modular landscape of the Human Infectome Network. The deconvolution of HIN using the CFinder algorithm identified interconnected modules of proteins (nodes) and modules' linkers (available in an interactive format in Additional file 1). Protein modules and linkers are coloured according to the intensity of the viral attack. Highly targeted modules or linkers are red. Poorly targeted modules or linkers are black. Biological processes and molecular functions associated to highly targeted modules are pinpointed (one-tailed Exact Fisher test; Benjamini and Hochberg multiple testing correction; P-Value < 0.05 red arrows - P-Value < 0.15 grey arrows). c. Simulation of network robustness against preferential viral attack on central and bridging proteins. The figure represents the fragmentation of the entire human protein interaction network (HPIN) according to random or preferential attack according to network properties. The fragmentation is obtained by computing the relative size of the largest connected component (S) as a function of the percentage of removed nodes (f). Nodes removal is performed either randomly (black) or as a preferential attack mode where protein nodes are eliminated from the network in decreasing order of their connectivity (blue), centrality (red) and bridging (pink) values.

Figure 3

Cellular connectivity (k_h) and centrality (b_h) of Essential Host Factors (EHFs). The histograms show the average cellular connectivity (left - k_h) and centrality (right - b_h) of Essential Host Factors (EHFs) computed for the group of proteins directly interacting with viruses (left; red bars - TP) and for the groups of proteins indirectly targeted by viruses, (right; orange bars - TP-N). Values are compared to average values of connectivity and centrality computed for a control dataset of not-Essential Host Factors (NOT-EHFs). Distribution differences are statistically assessed using one-tailed Wilcoxon test (P-Values < 0.05 *).

Figure 4

The Human Infectome-Diseasome Network (HIDN) . a. Diseasome classification of Targeted Proteins. Viruses interact directly with Disease-Related Proteins (DRPs) but also indirectly throughout 1-hop interaction. The distribution of directly targeted DRPs (TPs, red) or indirectly targeted DRPs (TP-Ns, orange) is given. b. Bridging properties of Disease-Related Proteins. The histograms show average cellular bridging properties (br_h) of DRPs for proteins directly targeted by viruses (left; red bars - TPs) and for proteins indirectly targeted by viruses, at a 1-hop distance, (right; blue bars - TP-Ns). These average values are compared to average values of bridging computed for a control dataset of not-diseases related proteins (NOT-DRPs). Differences are statistically assessed using one-tailed Wilcoxon test (P-Value < 0.05 *; P-Value < 0.001 ***). c. OMIM enrichment analysis of the proteins targeted directly and indirectly by viruses. The table shows 34 OMIM diseases significantly connected to viruses both directly and indirectly after multiple testing corrections. For each significant disease, OMIM id, description, disease type, number of targeted proteins, fold enrichment value and the Benjaminin and Hochberg adjusted P-Value are given. d. The Human Infectome-Diseasome Network (HIDN). The HIDN was mathematically formalized as a bipartite graph composed of two types of nodes corresponding to either diseases (black nodes) or viruses (coloured nodes). HIDN is composed of 57 viruses and 230 diseases connected by 466 virus-disease associations. In HIDN, disease and viral nodes are connected by an edge if at least one protein related to this disease is targeted by at least one protein encoded by this virus. Viral species and disease are connected by an edge if at least one disease related protein associated to that disease is directly targeted by a protein encoded by the virus. The nodes are sized proportionally to disease connectivity (k_d) or virus species connectivity (k_vs) in HIDN.

Figure 5

The Hepatitis C Virus Infectome Diseasome Network. a. The Hepatitis C Virus Infectome-Diseasome Network. The network is represented as a multi-coloured graph with three types of node: viral proteins (red circle), host cellular proteins (blue circle) and diseases (black circles). Virus-host protein-protein interactions and disease-gene associations are respectively represented by red and black edges. b. Hierarchical clustering of Hepatitis C Virus proteins according to their connectivity to human diseases. The closeness index, i.e. the reciprocal of the average distance between viral proteins and diseases were computed within neighbourhood-based HIDN, was used as a distance metric for unsupervised hierarchical clustering. c. HIDN connectivity of HCV proteins and main HCV-associated diseases: Hepatocellular carcinoma (left) and Cirrhosis (right).

Figure 6

The Infectome-Autoimmune Diseasome Network. The Infectome-Autoimmune Diseasome Network is modelled as a multi-coloured graph with two types of nodes (diseases - black circles and host cellular proteins - blue square). Host cellular proteins can be either directly connected to the disease (DRPs) or indirectly connected through 1-hop distance (DRP-Ns). Protein-protein interactions between host cellular proteins are represented by blue edges. Disease-gene associations are represented by black edges. DRPs targeted by viruses are represented with red stroke colour (IFIH1, OAS1). DRP-Ns targeted by viruses are represented with orange stroke colour (VISA).