Networks have also been used for the study of somatic mutations occurring in metastatic melanoma. In a recent study, a large protein interaction network was used to find sub-clusters or modules of interacting proteins that were PDGFR inhibitor affected in tumors. Whilst the genes affected by somatic copy number variants were different in different tumors, they often occurred in the same modules of proteins, which were in turn associated with cell cycle and apoptotic functions [88]. These two examples used biological networks composed of known protein interaction and pathway data, and mapped genetic observations

to these networks. An alternative approach is to generate a network from the data itself, rather than from additional functional information. The advantage of this approach is that the network reflects the data of a specific controlled experiment rather than data from

many different experiments, often from many different cell types. Because the network does not rely on known relationships, observations made in such networks can lead to truly novel discoveries. A recent example of such a study used global gene expression profiles from human pancreatic islets and identified a network module containing Sfrp4, which was strongly over-expressed in non-insulin-dependent diabetes mellitus patients and affected insulin secretion [85], [89] and [90]. Network theory has shown that the most connected genes within biological networks (the hub genes) TSA HDAC are often the most essential [76]. In the abovementioned study, Sfrp4 was identified as a hub gene in the module, and was as

such identified as an important putative target affecting insulin secretion. The identification of this gene would not have been possible without looking at the interconnectedness of the genes in the context of all the experimental data. Considering networks Methocarbamol of pathways (instead of single gene products) as being affected comparing 2 phenotypes is particularly adapted to the dissection of fine metabolic modulations, particularly in experimental settings associated with high biological variation [91], as with human samples. Moreover, network biology better reflects the physiological situation–where the modulation of a given molecule of interest affects many different factors–topologically visible as clusters (Fig. 6). This integration allows the exploitation of the complementary aspects of different data sets, going one step further than simply considering common gene product regulation among mRNA and proteins. Known protein–protein interactions and pathway database information can also be used to weight experimental relationships and complement the network. Then, interpretation of the network can be performed using gene-set or gene-ontology enrichment analysis [92], or other bioinformatics tools [93]. Finally, validation of such results can be performed in vivo or using biological models, reproducing the same phenotype by modulating the pathway of interest [74].