Publication Type:Journal Article
Source:Nephrol Dial Transplant, Volume 27, Issue 11, p.3995-4002 (2012)
Keywords:asb, Biomarkers, Computational Biology, Diabetic Nephropathies, Humans, Metabolomics, Models, Biological, Proteomics, Systems Biology
The pathophysiology of diabetic nephropathy (DN) is driven by a complex, multi-facetted interplay of numerous molecular processes (protective as well as damaging) and the balance between these, rather than the activity of a single pathway, determines clinical presentation and outcome. We present a concept for deriving a biomarker panel aimed to represent the relevant processes involved. Our approach rests on a hybrid gene/protein interaction network that holds ample information on molecular features (nodes) and their relations (edges), as a result providing a basic structure to navigate in molecular content and context being identified as relevant in DN. Extensive literature search on omics studies in DN provided a molecular feature list mapping to a total of 2175 unique protein-coding genes [13 from single nucleotide polymorphisms (SNPs), 12 as targets from relevant miRNAs, 1583 from transcriptomics, 5 from proteomics and 53 from metabolomics via linking to enzymes; 509 features were identified from multiple sources]. Two hundred and eighty-seven further human protein-coding genes associated with DN were derived from searching NCBI Pubmed (utilizing MeSH and gene-to-pubmed). Text mining of patents and clinical trial descriptors in the context of DN further added about 1,000 features. These data were used to label the respective nodes in the interaction network, as a result obtaining a DN-specific subgraph. Application of a segmentation algorithm on this subgraph allowed the identification of DN-specific molecular units, each characterizing a cluster of genes/proteins with a high internal functional association. We interpret each such unit as a functionally relevant molecular process contributing to the presentation of DN, and the total set of such units as a molecular model of DN. We propose that selecting appropriate biomarkers from each unit might allow the description of a patient's specific 'type' of DN, ultimately leading to a better stratification of patients regarding progression risk and optimal interventional approach.