Characterization and comparison of gene-centered human interactomes

Abstract

The complex web of macromolecular interactions occurring within cells-the interactome-is the backbone of an increasing number of studies, but a clear consensus on the exact structure of this network is still lacking. Different genome-scale maps of human interactome have been obtained through several experimental techniques and functional analyses. Moreover, these maps can be enriched through literature-mining approaches, and different combinations of various 'source' databases have been used in the literature. It is therefore unclear to which extent the various interactomes yield similar results when used in the context of interactome-based approaches in network biology. We compared a comprehensive list of human interactomes on the basis of topology, protein complexes, molecular pathways, pathway cross-talk and disease gene prediction. In a general context of relevant heterogeneity, our study provides a series of qualitative and quantitative parameters that describe the state of the art of human interactomes and guidelines for selecting interactomes in future applications.

Keywords: comparison; interactome; molecular interactions; network biology.

Figure 1

Characterization of human interactomes. We considered topological properties, protein complexes, molecular pathways, PCT and performance in disease gene prediction. Our study describes the state of the art and offers a series of hints to guide the choice of interactomes in future applications.

Figure 2

Topological properties of 19 human interactomes. (A) Number of interactions versus number of genes (size). (B) Density versus size. (C) Diameter versus mean distance; dot size is proportional to the mean gene transitivity (or clustering coefficient, i.e. the fraction of closed triangles in the network). (D) Overlap between genes, defined as the ratio between the genes shared by each couple of interactomes and the size of the interactome in the corresponding column label; this implies that a row indicates to which extent the interactome (row label) includes other interactomes, while a column indicates to which extent the interactome (column label) is included in other interactomes. (E) Interaction overlap, defined analogously to gene overlap. (F) Assessment of the scale-free hypothesis: power law exponent (alpha) and P-value; circles (exponential): the exponential distribution fits better than power law; triangles (none): power law fits better than other distributions. (G) Average correlation values of topological measures; the dendrogram was obtained by complete linkage method.

Figure 3

Protein complexes. (A) Average CCF (<CCF>) and standard deviation of CCF (SD(CCF)) across interactomes. (B) CCFs of three protein complexes: HOOK2-KCL3-LRGUK1-RIMBP3 (HKLR), GPI-GnT (GG) activity complex and Arp2/3 (Arp) complex. (C) Network visualization of three protein complexes with link colored by their occurrence in the interactomes. (D) Number of protein complexes (#) with CCF > 0.5 in relation to interactome size. (E) Heatmap of CCF values. (F) Average correlation values of protein complex CCFs; the dendrogram was obtained by complete linkage method.

Figure 4

Molecular pathways. (A) Average CCF (<CCF>) and standard deviation of CCF (SD(CCF)) across interactomes; big dots represent median values calculated for each pathway category. (B) CCFs of three molecular pathways: nitrogen metabolism (NIT), synthesis and degradation of ketone bodies (KET) and DNA replication (DNAR). (C) Network visualization of three molecular pathways with link colored by their occurrence in the interactomes. (D) Number of molecular pathways (#) with CCF > 0.5 in relation to interactome size. (E) Heatmap of CCF values. (F) Average correlation values of pathway CCFs; the dendrogram was obtained by complete linkage method.

Figure 5

PCT. (A) Network representation of four pathways in GH interactome: GG, glycolysis/gluconeogenesis; AAG, alanine, aspartate and glutamate metabolism; HEPA, glycosaminoglycan biosynthesis—heparan sulfate/heparin; GPI, glycosylphosphatidylinositol GPI-anchor biosynthesis; only the interactions among the genes belonging to such pathways are shown. (B) Network proximity (X*) of genes of pathways AAG, HEPA and GPI from genes of pathway GG in DMN interactome. (C) PCT between GG–AAG, GG–HEPA and GG–GPI. (D) Average (<PCT>) and SNR of PCT between all-pairs pathways across interactomes. p53, p53 signalling; PC, pancreatic cancer. (E) Correlations (Spearman) of PCTs among interactomes. (F) Correlation values of PCTs; the dendrogram was obtained by complete linkage method.

Figure 6

Disease gene prioritization. (A) Average correlation values between gene prioritizations. (B) Overlap of top ranking genes between interactomes. (C) Correlation of disease prioritization results between interactomes by disease. (D) Overlap of the top ranking genes between interactomes by disease. (E) Performance of each interactome estimated by means of 5-fold cross-validation; for clarity, lines have been added between the points; interactomes are ordered by decreasing average performance over all diseases considered (from right to left). (F) Average performance over the five diseases in relation to interactome size. (A, B) Dendrograms were obtained by complete linkage method.

Figure 7

Overall interactome similarity. (A) Overall similarity network among interactomes; for each interactome, the arrows point to its four most similar interactomes (see Methods); yellow colored areas indicate community structure. (B) Aggregate correlation (R), decomposed in the contributions of each analysis type.