Modelling human protein interaction networks as metric spaces has potential in disease research and drug target discovery

Abstract

BackgroundWe have recently shown by formally modelling human protein interaction networks (PINs) as metric spaces and classified proteins into zones based on their distance from the topological centre that hub proteins are primarily centrally located. We also showed that zones closest to the network centre are enriched for critically important proteins and are also functionally very specialised for specific ‘house keeping’ functions. We proposed that proteins closest to the network centre may present good therapeutic targets. Here, we present multiple pieces of novel functional evidence that provides strong support for this hypothesis.ResultsWe found that the human PINs has a highly connected signalling core, with the majority of proteins involved in signalling located in the two zones closest to the topological centre. The majority of essential, disease related, tumour suppressor, oncogenic and approved drug target proteins were found to be centrally located. Similarly, the majority of proteins consistently expressed in 13 types of cancer are also predominantly located in zones closest to the centre. Proteins from zones 1 and 2 were also found to comprise the majority of proteins in key KEGG pathways such as MAPK-signalling, the cell cycle, apoptosis and also pathways in cancer, with very similar patterns seen in pathways that lead to cancers such as melanoma and glioma, and non-neoplastic diseases such as measles, inflammatory bowel disease and Alzheimer’s disease.ConclusionsBased on the diversity of evidence uncovered, we propose that when considered holistically, proteins located centrally in the human PINs that also have similar functions to existing drug targets are good candidate targets for novel therapeutics. Similarly, since disease pathways are dominated by centrally located proteins, candidates shortlisted in genome scale disease studies can be further prioritized and contextualised based on whether they occupy central positions in the human PINs.

Self-similarity of human protein interaction networks: a novel strategy of distinguishing proteins

“…The starting point is that in all the networks under consideration, the centres were identified, and nodes were grouped (in zones) according to the distances they are from these central proteins. With this classification, functional enrichment was performed and biological hypotheses were drawn 9 25 .…”

“…It has recently been observed that there is some level of specialization by proteins in various zones of the HFPIN 25 . Also, while some pathways cut across zones, of importance is that sensing pathways are far more pronounced in central zones than in periphery.…”

“…In our recent work when we considered the distribution of proteins that consistently expressed in 13 types of cancer 25 , it was shown that most of these proteins are prominent in zone 2 of the HFPIN, HSN and CHN ( tables 7 to 9 ). Here, the same methods were applied as we analysed each of the subgraphs from each zone.…”

“…Second, it was observed that some zones are uniquely-enriched and represent a far more pronounced specialisation. Third, it was shown that cancer pathways are significantly over represented in zone 2 25 .…”

Clinical implications of omics and systems medicine: focus on predictive and individualized treatment

“…This property may be helpful when prioritizing between multiple genes in a disease module for functional and clinical studies, as well as when selecting genes as potential drug targets. It is interesting that genes targeted by drugs are more likely to be hubs . The network property that highly interconnected nodes are likely to be functionally related can be exploited to find novel diagnostic markers and therapeutic targets amongst the interactors of known disease genes .…”